Assisted Improvement of Security Reliance Scores

Abstract

In representative embodiments, a system and method to recommend improvements to a security reliance score is illustrated. Individual cryptographic key material has an associated security reliance score that is calculated based on attributes of associated with the cryptographic key material. The system identifies an improvement goal and evaluates a selected cross-section of key material and their associated scores. Based on the evaluation, the system creates an exemplary model having attributes to use as the basis of improvement. This model is then used to calculate improvement potential for a selected cross-section of scores by calculating the improvement that would occur if changes were made in the selected cross-section to match the model. Based on the improvement potential, the system can then automatically initiate action(s) to improve scores or present options for action(s) to a user for selection and initiation.

Description

CROSS RELATED TO RELATED APPLICATIONS

This application is a Continuation in Part of U.S. patent application Ser. No. 14/802,502, filed Jul. 17, 2015, which application also claims the benefit of U.S. Provisional Application 62/025,857, filed Jul. 17, 2014. This application also claims the benefit of U.S. Provisional Application 62/279,447, filed Jan. 15, 2016. All of these applications are incorporated by reference herein their entirety.

FIELD

This application relates generally to improving security in computer systems and networks and more specifically to systems and methods that can assist and guide users to improve security reliance score for cryptographic key material, protocols, systems and process configurations used in such systems.

BACKGROUND

Assessing vulnerabilities related to and the credibility of cryptographic material in systems is a difficult problem. Many solutions attempt to evaluate various aspects of a system or protocols in isolation to identify whether vulnerabilities exist. However, aspects evaluated in isolation do not always provide a good understanding of the security or trustworthiness of a system or process. Furthermore, identifying how to improving security and/or trustworthiness is often extremely difficult given the various factors that may influence security and/or trustworthiness.

BRIEF DESCRIPTION OF DRAWINGS

FIG. 1 illustrates an example diagram for calculating a security reliance score.

FIG. 2 illustrates an example flow diagram for calculating a security reliance score.

FIG. 3 illustrates an example deployment architecture.

FIG. 4 illustrates another example deployment architecture.

FIG. 5 illustrates representative details of the flow diagram of FIG. 2.

FIG. 6 illustrates a representative function for calculating an update vector.

FIG. 7 illustrates a representative function for calculating an anomaly score.

FIG. 8 illustrates a representative vulnerability scale.

FIG. 9 illustrates a representative user interface for a security reliance score improvement recommendation system.

FIG. 10 illustrates a flow diagram detailing operation of a security reliance improvement system.

FIG. 11 illustrates a flow diagram for creating a model for use in making security reliance core improvement recommendations.

FIG. 12 illustrates a another flow diagram for creating a model for use in making security reliance core improvement recommendations.

FIG. 13 illustrates a flow diagram for identifying actions to present to a user.

FIG. 14 illustrates a representative software architecture.

FIG. 15 is a block diagram of a machine in the example form of a processing system within which may be executed a set of instructions for causing the machine to perform any one or more of the methodologies discussed herein including the functions, systems and flow diagrams thereof.

DETAILED DESCRIPTION

The description that follows includes illustrative systems, methods, user interfaces, techniques, instruction sequences, and computing machine program products that exemplify illustrative embodiments. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide an understanding of various embodiments of the inventive subject matter. It will be evident, however, to those skilled in the art that embodiments of the inventive subject matter may be practiced without these specific details. In general, well-known instruction instances, protocols, structures, and techniques have not been shown in detail.

OVERVIEW

This disclosure describes systems and methods to help assess the security reliance, trustworthiness, and reliability of cryptographic material. The disclosure also describes systems and methods to help improve the security reliance, trustworthiness, and reliability of cryptographic material, protocols, systems and process configurations. In one embodiment, a system gathers information from a system, group of systems, company and so forth and uses the information to calculate a security reliance score based on the cryptographic material and the context in which it is used. Collection and consideration of such a large body of data cannot be performed by a human and allows the system to evaluate some unique aspects of both the cryptographic material and the context in which it is used that are simply not possible in more manual evaluations. Furthermore, the system employs learning models, statistical analysis and other aspects that simultaneously account for an ever changing environment and produce results that are not possible when similar data is manually evaluated. As used herein, cryptographic material is a broad term used to encompass material used in a security context and includes material used with a cryptographic algorithm such as cryptographic keys, certificates and so forth.

In some embodiments, the security reliance score can be used as an indication of the vulnerability of systems and protocols applying the evaluated cryptographic material. To help with this, the security reliance score is mapped to a vulnerability scale in some embodiments. The score's metric accounts for various factors, including weighted, autonomous or interdependent factors such as known vulnerabilities; compliance to standards, policies, and best practices; geographic locations and boundaries; and normative deviations through statistical analysis, extrapolation, and heuristic contingencies. In some embodiments, the scoring is further dynamically adjusted to identify the trustworthiness of a particular system, its cryptographic material, and the usage of its cryptographic material in response to learned patterns in incoming data and a dynamic and ever changing environment.

Security reliance scores are calculated by evaluating various properties and attributes of cryptographic material and the context in which the cryptographic material is used. Individual scores for attributes can be aggregated into a property score and property scores can be aggregated into a security reliance overall score for the cryptographic material under consideration. Scores for cryptographic material can be further aggregated to evaluate an overall system, cluster of systems, site, subsidiary, company, vertical and so forth.

Initial values for the scores are determined and algorithms employed that modify the scores over time based on various factors and changes that occur. Learning algorithms, pattern recognition algorithms, statistical sampling methods and so forth are employed in various embodiments as outlined in greater detail below.

Security reliance scores can be used in a variety of contexts. In one embodiment, security reliance scores are used by others to determine whether and to what extent to trust a system or other entity. In other embodiments, the system can identify which of the various factors used in generating the security reliance score would have the most impact on the security reliance score, thus assisting and directing administrators or others striving for an improvement to evaluate the impact of changes within a system, site, company and so forth. In still other embodiments, security reliance scores from different entities can be compared to determine a relative accepted normative baseline. For example, companies within a vertical industry can be compared to ascertain compliance with a normative minimum accepted standard amongst peers and to identify positive and negative outliers form such norm. Other uses for the security reliance scores also exist.

A recommendation system to improve security reliance scores associated with cryptographic material is also disclosed. As described herein, security reliance scores can be at different levels and aggregated across levels all the way to an ultimate aggregated score. For example, all the cryptographic scores for the cryptographic material in an entire company can be aggregated for to arrive at a security reliance score for the company. Examples of different levels that can be aggregated include, but are not limited to, an industry, company, subsidiary, location, facility, system, group(s) of systems, and so forth.

In the context of a recommendation system, some cross section of cryptographic material of interest and their attendant scores can be evaluated in order to help administrators, users, or others that have responsibility for a cross section of the cryptographic material improve the associated scores and thus improve the security reliance, trustworthiness, and/or reliability of the underlying cryptographic key material, protocols, systems, process configurations and so forth. This evaluation can be interactive in nature or can be automatic in response to the occurrence of one or more events.

Security reliance scores of interest (i.e., some cross section of scores that are to be evaluated) are retrieved. If the system is used in an interactive manner, statistics and/or metrics associated with the scores can be presented to a user, administrator, or other individual who has authority to evaluate the scores. In the following description, the term “user” will be used to encompass any of these. Statistics and/or metrics can be calculated in a variety of ways. In one example, the scores falling into designated percentile ranges are identified and the number of scores falling into each designated percentile range is aggregated in order to identify what percentage of the scores fall into each of the designated percentile ranges. The aggregated scores along with the overall aggregated scores at different levels can be presented, if desired. This allows an individual exploring an aggregate score to evaluate various combinations and identify specific targets for improvement. Improvement can be associated with one or more ‘goals’ such as increasing the percentage of scores that are “good,” e.g., expressed as a range of percentiles with desirable scores, reducing the percentage of scores that are “bad,” e.g., expressed as a range of percentiles with undesirable scores, improving the aggregate score at a particular level, or similar. As discussed in detail herein, a goal comprises a primary improvement metric and a possible secondary improvement metric.

The system calculates improvement potential based on the attributes of the underlying cryptographic material as compared to a model set of attributes derived from a subset of the selected cross-section. Potential actions consistent with improvement goals are identified and presented to the user. The system can also provide interactive information to show the impact of selected courses of action. In a non-interactive implementation, the system can calculate improvement potential and select one or more actions to implement.

ACRONYM GLOSSARY

The following is an acronym glossary along with relevant specifications that define and/or discuss the associated acronym definition, as appropriate.

• • 2TDEA Two-key Triple Data Encryption Algorithm (NIST SP-800-57, Part I)
  • 3TDEA Three-key Triple Data Encryption Algorithm (NIST SP-800-57, Part I)
  • AES Advanced Encryption Standard (FIPS 197)
  • AIA Authority Information Access (RFC 5280)
  • ANSI American National Standards Institute
  • BGP Border Gateway Protocol (RFC 4271)
  • CA Certification Authority (NIST SP-800-57, Part I, Glossary)
  • CBC Cipher Block Chaining (NIST SP-800-38A)
  • CDP CRL Distribution Point (RFC 5280)
  • CRL Certificate Revocation List (RFC 5280)
  • DANE DNS-based Authentication of Named Entities (RFC 6698)
  • DH The FFC Diffie-Hellman key-agreement primitive (NIST SP-800-56A Revision 2, Glossary).
  • DNSSEC Domain Name System Security Extensions (RFC 4033)
  • DSA Digital Signature Algorithm (FIPS 186-3)
  • DV Domain-vetted X.509 TLS server certificates.
  • EC Elliptic Curve (NIST SP-800-56A Revision 2, Glossary).
  • ECC Elliptic Curve Cryptography, the public-key cryptographic methods using operations in an elliptic curve group (NIST SP-800-56A Revision 2, Glossary).
  • ECDH The ECC Diffie-Hellman key-agreement primitive.
  • ECDHE ECDH based on an ephemeral key pair. An ephemeral key pair is a key pair, consisting of a public key (i.e., an ephemeral public key) and a private key (i.e., an ephemeral private key) that is intended for a short period of use (NIST SP-800-56A Revision 2, Glossary).
  • ENISA European Network and Information Security Agency
  • EV X.509 TLS server certificates complying with “Guidelines For The Issuance And Management Of Extended Validation Certificates, v.1.5.5,”, 2015, CA/Browser Forum,
  • FFC Finite Field Cryptography, the public-key cryptographic methods using operations in a multiplicative group of a finite field (NIST SP-800-56A Revision 2, Glossary)
  • FIPS Federal Information Processing Standards Publications
  • GCM Galois/Counter Mode (NIST SP-800-38D)
  • HMAC Keyed-Hash Message Authentication Code (FIPS 198)
  • HSTS HTTP Strict Transport Security (RFC 6797)
  • IETF Internet Engineering Task Force.
  • IPSec IP Security (RFC 4301)
  • ITU International Telecommunication Union
  • MD5 Message-Digest algorithm 5 (RFC 1321)
  • NIST National Institute of Standards and Technology
  • OCSP Online Certificate Status Protocol (RFC 6960)
  • OV Organization-vetted X.509 TLS server certificates.
  • PFS Perfect Forward Secrecy
  • PM Public-Key Infrastructure (NIST SP-800-57, Part I, Glossary)
  • RFC Request for comment, see http://www.ietforg/rfc.html. RFCs are identified by a number, such as RFC 4346 or RFC 6066.
  • RSA Rivest, Shamir, Adelman (an algorithm) (R. L. Rivest, A. Shamir, and L. Adleman, “A Method for Obtaining Digital Signatures and Public-key Cryptosystems,” Communications of the ACM, 21, 1978, ACM, pp. 120-126.)
  • S-BGP Secure Border Gateway Protocol (Seo, K.; Lynn, C.; Kent, S., “Public-key infrastructure for the Secure Border Gateway Protocol (S-BGP),” DARPA Information Survivability Conference & Exposition II, 2001. DISCEX '01. Proceedings, vol. 1, pp.239-253)
  • SCT Signed Certificate Timestamp (RFC 6962)
  • SHA-1 Secure Hash Algorithm (FIPS 180-3)
  • SHA-256 Secure Hash Algorithm (FIPS 180-3)
  • SSH Secure Shell (RFC 4251)
  • SSL Secure Socket Layer (RFC 6101)
  • TLS Transport Layer Security (RFC 5246)
  • TLSA DANE resource record (RFC 6698)
  • X.509 ITU X.509.

DESCRIPTION

Embodiments comprise a security reliance metric for assessing cryptographic material based on a variety of weighted, independent, or interdependent factors, such as known vulnerabilities; compliance to standards, policies, and best practices; geographic locations and boundaries; and normative deviations through statistical analysis and extrapolation, and heuristic contingencies. Some embodiments dynamically adjust initial empirical scoring assignments based on learning patterns.

When assessing the security reliance of cryptographic material, various factors, either independent or correlated, impact the overall security reliance. When considering cryptographic material, the security reliance factors can be broadly broken down into factors relating to the cryptographic material itself and factors related to the protocol, context or other environment in which it is used. Throughout this disclosure TLS will be used as an example although the principles of the disclosure equally apply to any type of cryptographic material such as public/private keys used in SSH, IPSec, S-BGP, and DNSSEC. The following presents a simple overview of TLS as an example as context for the disclosure.

One commonly applied workflow for TLS uses X.509 certificates to establish a secure and authenticated connection between two systems. Thus, TLS uses both cryptographic material (the X.509 certificate) and a protocol (TLS) to establish the secure connection.

FIG. 1 illustrates a conceptual system architecture 100 for determining a security reliance score 112 for assessing cryptographic material. As explained in more detail below, a security reliance score 112 is based on (block 102) a plurality of property scores (108, 110). As explained in further detail below, in some embodiments the security reliance score 112 is a weighted aggregation 102 of individual property scores (108, 110). Properties scored for particular cryptographic material typically include properties for the cryptographic material itself and/or the environment or context in which the cryptographic material is used. Using TLS as an example, properties may include, but are not limited to one or more properties for X.509 certificate (or other cryptographic material) and one or more properties for the TLS configuration.

As further explained below in some embodiments, property scores (108, 110) are determined and/or calculated using specific aggregating functions (104, 106) having as inputs individual attribute scores (114, 116, 118, 120) that make up the properties. These specific aggregating functions can be selected based on the attributes. In the embodiments shown below, the aggregating functions in one case is a weighted sum. In another case, the aggregating function is a table lookup that takes as an input individual attribute scores and produces as an output the property score. In yet another case, the function is an assignment of a score based on some attribute value (like estimated security strength). In yet another case, individual attribute scores are used as input into a table lookup and the resultant values from the table used as input into a weighted sum. In the representative embodiments below, these aggregating functions are chosen to illustrate the variety of aggregating functions that are possible. Furthermore, it illustrates the principle that some types of attributes lend themselves more closely to a particular type of aggregating function than other types of aggregating functions.

Again using TLS as an example, attributes that make up the X.509 certificate property and TLS configuration property may include, but are not limited to:

1. Example X.509 certificate (or other cryptographic material properties):

• • i. Public key length;
  • ii. Public key algorithm;
  • iii. Certificate's validity period;
  • iv. Public key's cryptoperiod;
  • v. Certificate's signature algorithm;
  • vi. Certificate revocation status repository references such as CDP and OCSP (Authority Information Access);
  • vii. Configuration of certificate extension attributes, e.g., key usage, extended key usage, policy and name constraints.
  • viii. Certificate's issuer vetting process (DV, OV, EV);
  • ix. Certificate's issuer origin;

2. Example TLS configuration attributes:

• • i. TLS compression enabled/disabled;
  • ii. (Multiple-) Certificate status request enabled/disabled;
  • iii. TLS insecure renegotiation enabled/disabled;
  • iv. Protocol version support (best and worst);
  • v. Cipher suite support (best and worst) and cryptographic primitives configuration, e.g., PFS support, block-cipher chaining mode, block-cipher authentication mode;
  • vi. Session resumption support and implementation, e.g., session tickets as described in RFC 5077.

As indicated by adjustment operations (142, 144, 146, 148, 150, 152, 154) the various scores can be adjusted by a variety of functions. The adjustment operations are illustrated as optional as not all embodiments need employ such adjustments. The adjustment operations are also optional in that in the embodiments that do employ adjustments, not all attribute scores, property scores, or security reliance score are adjusted. In the representative embodiments below, learning algorithms, pattern recognition and statistical sampling are used to adjust one or more attribute scores and the security reliance score. The former based on changes in environment over time and the latter based on whether the cryptographic material/environment are anomalous in some fashion. The machine learning algorithms, pattern recognition, statistical sampling, and/or other analytical algorithms are represented by analytics 156, which drives the adjustments (142, 144, 146, 148, 150, 152, 154). Not all adjustments use the same algorithms or methods of calculation and the representative embodiments below show such variations.

Weight operations (130, 132, 134, 136, 138, 140) illustrate that the attribute and/or property scores can be weighted in some instances (possibly after adjustment). For example, if the aggregating function (104, 106, and/or 102) is a weighted sum, the weight operations (130, 132, 134, 136, 138, 140) can represent the individual weights applied to the attribute and/or property scores (as appropriate) before summing.

Summarizing the above discussion, individual attribute values (122, 124, 126, 128) are optionally adjusted (142, 144, 146, 148), optionally weighted (130, 132, 134, 136) and aggregated (104, 106) to produce property scores. These property scores are, in turn, optionally adjusted (150, 152) and optionally weighted (138, 140) to produce property scores (108, 110) which are further aggregated (102) to produce a security reliance score (112), which again may be adjusted (154).

Although not illustrated in the diagram, individual security reliance scores 112 can be further aggregated using the same structure (e.g., optionally adjusted and/or optionally weighted values of security reliance values further aggregated to provide higher level security reliance scores, which are further aggregated and so forth) to produce security reliance scores for systems, groups of systems, cryptographic material holders, company regions, subsidiaries, and so forth to produce security reliance scores at multiple levels throughout a company, geographic region, vertical industry, or any other categorization. In these further aggregations, weighed sums, averages, lookup tables, and so forth can all be utilized in this further aggregation.

In one representative embodiment, further aggregations are done on a system, business line, enterprise and business vertical level. System can include either individual systems or collections of systems, like a data center or other collection. Business line includes departments or functions within an enterprise, such as accounting, legal, and so forth. Enterprise includes either a major component of an enterprise (subsidiary, country operations, regional operations, and so forth), or then entire global enterprise. A business vertical includes either the business or major components categorized into a standard category representing the type or area of business, such as the Global Industry Classification Standard (GICS) used by MSCI, Inc. and Standard & Poor's.

In order to perform the aggregation of security reliance scores on these various levels, aggregating functions can be used. In one example embodiment an average of security reliance scores from cryptographic material at the relevant levels is used as the aggregate security reliance score for that level. In another example embodiment, in order not to have low security reliance scores balanced out by high security reliance scores, security reliance scores can be used to identify a customizable number of configurations that meet a designated criteria. In one embodiment, the configurations with the 10 lowest security reliance scores are identified. These configurations can then be compared to peer configurations at the system, business line, enterprise and/or business vertical level to compare aggregate security reliance across these various levels.

FIG. 2 illustrates a representative flow diagram 200 illustrating processing algorithms associated with calculating security reliance scores. As explained below, the process takes initial values and then applies learning algorithms, pattern matching, statistical analysis, surveys, and other information to constantly update the security reliance scores to account for a shifting security environment and to ensure that the security reliance scores reflect the current reality.

The process starts at 202 and proceeds to operation 204 where the initial attribute and/or property values are identified and set. Although not explicitly shown, identifying which set of attributes and/or properties are going to be utilized in the score can also be performed prior to setting the initial values.

Summary of the Scoring Model and Setting Initial Values

A methodology used in some embodiments to set the initial values of properties and attributes can rely on analytical work or heuristics previously performed offline. For example, publications exist that give estimates of security strength that can, in turn, be combined with other information using customizable or predefined rules in order to arrive at the initial values. In one representative example, “Recommendation for Key Management—Part 1: General (Revision 3)”, NIST Special Publication, 800-57, 2012, National Institute of Standards and Technology (hereinafter “Key Management Recommendations”), incorporated herein by reference describes a security strength measurement for particular key lengths. In embodiments illustrated below, information regarding security strength for various attributes from this and other sources are utilized along with heuristics to arrive at initial score mappings, as explained blow. As one example and as described in this reference, in 2015, the key length for an RSA key of 2048 bits corresponds to 112 bit security strength and indicates that by itself this can be considered as sufficient, though not optimal. Thus, a 0.8 particular initial value assignment for this attribute on a scale of [0,1] can account for a “sufficient, but not optimal” assessment. Throughout the disclosure values for properties and attributes will be illustrated on a scale of [0,1], and such values are used in some embodiments. However, other embodiments can use a different scale for values and all are encompassed within the disclosure.

Instead of assigning initial values based on the model that the various attributes are independent, correlation between several attributes are considered when assigning initial values in some embodiments. Such correlations can be identified either by offline analysis or through the learning algorithm (see below) employed in some embodiments. Correlations from the learning algorithm can be constantly adjusted leading to a dynamic score that accounts for a shifting security evaluation over time and thus, initial values can take into account the latest determination of correlation between attributes. For example, in the context of a TLS-secured connection, the key length of the public key embedded in an X.509 TLS server certificate and the validity period of such certificate based, for example, on determining the cryptoperiod of the underlying private key, are correlated. The Key Management Recommendations reference discussed above describes various attributes that can affect the cryptoperiod and suggests various cryptoperiods.

As one example, assume a X.509 certificate with an RSA public key of 2048 bits. As indicated above, in the absence of any correlation consideration, a value of 0.8 might be assigned as an initial value. When cryptoperiod is considered, however, the initial value may change. Continuing with the example, assume that a recommended cryptoperiod for a key of this type and this length is 1-2 years when the key is used for authentication or key exchange. If the certificate has a three-year validity period, the certificate deviates from the recommended cryptoperiod of 1-2 years for private keys used to provide authentication or key-exchange. To reflect this deviation, a 0.7 initial value can be assigned.

Development of the Scoring Model and Setting Initial Values

The following represents example embodiments of how initial scores are set in operation 204. As previously summarized in conjunction with FIG. 1, an overall score can be calculated as an aggregation of weighted property scores of security-relevant properties, P₀, . . . P_n, Such an aggregation takes the form of a weighted sum in some embodiments. Let P_i identify a property, W_Pi be a weight assigned to the respective property and σ_Pi be a scalar value representing the value of the property, whose calculation is described in detail below. The overall score, σ, can then be described as:

$\sigma \text{:=}\{\begin{array}{cc}\Delta ,& \mathrm{if}\mathrm{debasing}\mathrm{condition}\mathrm{is}\mathrm{met}\\ \Psi \left(\sum _{i=0}^n{\sigma }_{P_i}·W_{P_i},\Omega \right),& \mathrm{otherwise}\end{array}$

Where:

• • Δ is a constant scalar value representing a minimal customizable, but fixed value when one or more debasing conditions D₀, . . . , D_m is met. A debasing condition is a condition that would cause the security property to lose some or all of its value in terms of contribution to security strength. An example might be the discovery that a particular secret key has been compromised, e.g., by discovering the key on a seized hacking site. Debasing conditions can be defined as a list or rules that are updated periodically. A suitable value for A in some embodiments is 0.
  • σ_Pi is the value (score) for property P_i.
  • W_Pi is the weight for property P_i.
  • σ_i·W_Pi is the weighted score for property P_i and represents the initial value of the security reliance score (before any adjustment by the anomaly score).

Ω is an anomaly score derived from a statistical analysis, by applying dynamic pattern recognition and/or by evaluating additional context-sensitive data (described below).

• • Ψ:[0,1]×[0,1]→[0,1] is a function that aggregates the sum of the weighted property scores (Σ_i=0ⁿσ_Pi·W_Pi) and the anomaly score (Ω). This function is described below.

In the discussion that follows, all weights and values (σ) are assigned in the interval [0,1], although different intervals may be used for different embodiments. Each property P_i, for 0≦i≦n, comprises of a set of attributes A_0,Pi, . . . , A_k,Pi, describing specific configuration settings or other attributes, with a particular value, σ_Aj,Pi and a particular weight, W_Aj,Pi The property score σ_Pi for each property P_i is calculated based on a formula specific to the property. As described above in conjunction with FIG. 1, this can take the form of a sum of weighted attribute scores (e.g., P₀), as single score assignments (e.g., P₁), or as a lookup matrix of fixed attribute scores according to a property's attribute configuration (e.g., P₃) or some other way of combining the individual attribute scores into a property score.

As explained above, one method of assigning initial values is to utilize recommendations of relevant regulatory bodies like NIST to identify starting information (like configuration recommendations, security strength, etc.) and then select initial values, weights, and so forth based on heuristical assessment. For example, NIST provides in various publications recommendations on configurations, security strength (in bits) for cryptographic primitives, key lengths, cryptoperiods and so forth. These can be used, as shown below, to derive weights, scores and so forth.

In one embodiment assessing the cryptographic strength of TLS and related cryptographic material uses five properties:

• • 1. P₀ (TLS Configuration), which comprises configurable criteria addressing security relevant features of the TLS protocol;
  • 2. P₁ (TLS Security) comprises configurable security parameters;
  • 3. P₂ (Certificate Context) comprises security relevant infrastructure or application protocol configurations in which a X.509 TLS server certificate is being used;
  • 4. P₃ (Certificate Security) comprises of a certificate's security parameters; and
  • 5. P₄ (Revocation Infrastructure) comprises of the availability and accessibility of a certificate's relevant revocation infrastructure.

In an example embodiment, these properties might be weighted as follows: W_P0:=0.2, W_P1:=0.2, W_P2:=0.15, W_P3:=0.25, and W_P4:=0.2 respectively.

Calculation of the initial property scores P₀-P₄ will now be described for various embodiments.

In one embodiment, the property P₀ (TLS configuration) comprises three attributes: A_0,P0 (Compression); A_1,P0 ((Multiple) Certificate Status Request); and A_2,C0 (Renegotiation). The weights and attribute scores associated with the attributes in this embodiment are:

$W_{A_0,P_0}\text{:=}0.4,{\sigma }_{A_0,P_0}\text{:=}\{\begin{array}{cc}0.4,& \mathrm{TLS}\mathrm{Compression}\mathrm{enabled}\\ 1,& \mathrm{TLS}\mathrm{Compression}\mathrm{disabled}\end{array}W_{A_1,P_0}\text{:=}0.2,{\sigma }_{A_1,P_0}\text{:=}\{\begin{array}{cc}1,& (\left(\mathrm{Multiple}\right)\mathrm{Certificate}\mathrm{Status}\mathrm{Request}\mathrm{supported}\\ 0.6,& (\left(\mathrm{Multiple}\right)\mathrm{Certificate}\mathrm{Status}\mathrm{Request}\mathrm{not}\mathrm{supported}\end{array}W_{A_2,P_0}\text{:=}0.4,{\sigma }_{A_2,P_0}\text{:=}\{\begin{array}{cc}0.3,& \mathrm{TLS}\mathrm{Insecure}\mathrm{Renegotiation}\mathrm{enabled}\\ 1,& \mathrm{TLS}\mathrm{Insecure}\mathrm{Renegotiation}\mathrm{disabled}\end{array}$

A_0,P0 (Compression) refers to the TLS configuration option described in RFC 4346, Sec. 6.2.2, in which a compression algorithm other than CompressionMethod.null is chosen, A_1,P0 ((Multiple) Certificate Status Request) refers to RFC 6961 and RFC 6066, Sec. 8, A_2,P0 (Renegotiation) refers to the support of a vulnerable type of the insecure TLS renegotiation extension, see RFC 5746 for insecure and secure renegotiation.

Additionally, in one embodiment, debasing conditions are defined. D₀ (Certificate Expired):=Δ, D₁ (Certificate Revoked):=Δ might be considered as reasonable debasement conditions. Here D₀ defines the condition, in which the validity period of an investigated X.509 TLS server certificate is expired and D₁ the condition, in which an investigated X.509 TLS server certificate has been revoked by its issuing certification authority. If any of these two conditions is met by a X.509 TLS server certificate securing an investigated network service, the value Δ is assigned to the overall score. As indicated above, in some variations of this embodiment, Δ is zero, indicating that the debasing effect of an expired or revoked certificate cannot be compensated by any other security property configuration.

The scoring and weights of these attributes in this embodiment is based on known exploits or recommended best practices. An enabled TLS Compression leaves, for example, an HTTPS session susceptible to exploits targeted at TLS Compression, see RFC 7457 Sec. 2.6 and T. Polk, K. McKay, and S. Chokhani, “Guidelines for the Selection, Configuration, and Use of Transport Layer Security (TLS) Implementations”, NIST Special Publication, 800-52 Revision 1, 2014, National Institute of Standards and Technology (hereinafter “TLS Implementation Guidelines”), Sec. 3.7, for security considerations. Support for TLS's Certificate Status Request (the precursor to Multiple Certificate Status Request, which is recommended in TLS Implementation Guidelines, Sec. 3.4.2.4), is mandatorily required by NIST, (TLS Implementation Guidelines, Sec. 3.4.1.2) and when not supported, represents a deviation from recommended practice. An insecure TLS Renegotiation is susceptible to exploits; see RFC 7457 Sec. 2.10 and RFC 5746.

The property score σ_P0 for property P₀ in this embodiment might be calculated by summing up the weighted attribute score assignments of the attributes described above.

${\sigma }_{P_0}\text{:=}\sum _{j=0}^2W_{A_j,P_0}·{\sigma }_{A_j,P_0}$

In one embodiment the property P_i (TLS Security) might initially assign attribute scores empirically based on the strength of a cipher suite's cryptographic primitives, see RFC 5246 Appendix A.5 and Key Management Recommendations. In compliance with TLS Implementation Guidelines, Sec. 3.3.1, all cryptographic primitives are expected to provide at least 112 bits of security. With that background as a starting point, the attributes of P₁ are defined by different security strength (in bits) values, i.e., A_0,P1(<112), A_1,P1(112), A_2,P1(128), A_3,P1(192), and A_4,P1(256). The security strength values can be assigned initial attribute values of: σ_n,P1=0, σ_1,P1=0, σ_2,P1=0, σ_3,P1=0, and σ_4,P1=0.

The security strength of the weakest cryptographic primitive in the cipher suite, as defined in Key Management Recommendations, determines the attribute score assignment. In other words, the cryptographic primitives of a particular cipher suite are examined and the security strength of each cryptographic primitive is determined (e.g., by the values from Key Management Recommendations or in some other consistent fashion). The lowest relative security strength is then selected as the security strength associated with the cipher suite. Based on that security strength, the closest attribute value that does not exceed the actual security strength is selected and the corresponding score used for σ_AjP1. The property score, σ_P1, is then the selected score, σ_AjP1. Thus:

σ_P1:=σ_AjP1

• • where σ_AjP1 is the score corresponding to the value of the lowest strength security primitive.

As an example, if the lowest security strength of all the primitives of a particular cipher suite was 127 bits, then the attribute associated with the cipher suite would be A_1,P1 (112 bits) since 112 is the closest value that does not exceed the value of 127, and the attribute value of σ_A1P1:=0.6 would be assigned. As a more complicated example, consider the cipher suite defined by “TLS_RSA_WITH_AES_128_GCM_SHA₂₅₆.” This means that the cipher suite uses RSA for the key exchange, AES with a 128 bit key, Galois/Counter Mode (GCM) as the block cipher chaining mechanism, and a SHA-256 hashing algorithm. If the RSA key exchange is based on a public key size of at least 3072 bits, thus providing at least 128 bits of security strength, the cipher suite in this embodiment is assigned to the value A_2,P1, (128 bits), as AES-128 provides 128 bits of security strength (see Key Management Recommendations), even though SHA-256 for HMACs is considered to provide 256 bits of security strength (see Key Management Recommendations). An ephemeral DH key exchange, necessary in order to support Perfect Forward Secrecy (PFS), is similarly evaluated, e.g., an ECDHE key exchange based on the NIST approved curve P-256 is considered to provide 128 bits of security strength, see Key Management Recommendations and hence assigned to the value A_2,P1.

In one embodiment, the property P₂ (Certificate Context) comprises attributes declaring support for Certificate Transparency, see RFC 6962, support for DNS-Based Authentication of Named Entities (DANE), see RFC 6698, support for HTTP Strict Transport Security (HSTS), see RFC 6797, and support for Public Key Pinning Extension for HTTP (HPKP), see RFC 7469. The weights and attribute scores associated with the attributes in this embodiment are:

$W_{A_0,P_2}\text{:=}0.3,{\sigma }_{A_0,P_2}\text{:=}\{\begin{array}{cc}1,& \mathrm{SCT}\mathrm{present}\left(\mathrm{CT}\mathrm{supported}\right)\\ 0.6,& \mathrm{otherwise}\end{array}W_{A_1,P_2}\text{:=}0.1,{\sigma }_{A_1,P_2}\text{:=}\{\begin{array}{cc}1,& \mathrm{TLSA}\mathrm{resource}\mathrm{present}\left(\mathrm{DANE}\mathrm{supported}\right)\\ 0.8,& \mathrm{otherwise}\end{array}W_{A_2,P_2}\text{:=}0.3,{\sigma }_{A_2,P_2}\text{:=}\{\begin{array}{cc}1,& \mathrm{HSTS}\mathrm{supported}\\ 0.4,& \mathrm{otherwise}\end{array}W_{A_3,P_2}\text{:=}0.3,{\sigma }_{A_3,P_2}\text{:=}\{\begin{array}{cc}1,& \mathrm{HPKP}\mathrm{supported}\\ 0.6,& \mathrm{otherwise}\end{array}$

Similarly to property P₀ (TLS Configuration), the property score σ_P2 is again defined as the summation of the weighted attribute scores:

${\sigma }_{P_2}\text{:=}\sum _{j=0}^3W_{A_j,P_2}·{\sigma }_{A_j,P_2}$

In another embodiment attribute values might be correlated to a combination of conditions and/or other attributes in even different properties. In one embodiment a two-dimensional correlation can be represented by a matrix with a cell-based attribute score assignment. Assuming a uniform weight distribution, the property score can be retrieved by a table lookup in such matrix. If non-uniform weights are desired, after the table lookup, the property score can be weighted accordingly.

Calculating the scores for an example embodiment of P₃ (Certificate Security), illustrates such a correlation. This embodiment also illustrates an example where correlation between a combination of conditions attributed in different properties. In this embodiment, attributes of P₃ comprise the size of a public key embedded in a certificate (A_0,P3), its cryptoperiod (A_1,P3), whether PFS is supported and the key hashing algorithm used (A_2,P3). In this embodiment, the security strength (in bits) (see Key Management Recommendations) for the size of the public key embedded in a certificate is used to map the attribute A_0,P3 to an attribute score using the following mapping:

$A_{0_0,P_3}\left(\text{<}112\right),{\sigma }_{A_{0_0,P_3}}\text{:=}0,A_{0_1,P_3}\left(112\right),{\sigma }_{A_{0_1,P_3}}\text{:=}0.5,A_{0_2,P_3}\left(128\right),{\sigma }_{A_{0_2,P_3}}\text{:=}0.8,A_{0_3,P_3}\left(192\right),{\sigma }_{A_{0_3,P_3}}\text{:=}0.9,\mathrm{and}A_{0_4,P_3}\left(256\right),{\sigma }_{A_{0_4,P_3}}\text{:=}1$

The mapping is accomplished by selecting the attribute with security strength that is lower than, or equal to, the security strength of the corresponding key length.

A certificate's public key's cryptoperiod, attribute A_1,P3 is mapped to an attribute score using the following mapping (cryptoperiod measured in years):

$A_{1_0,P_3}\left(\text{>}5\right),{\sigma }_{A_{1_0,P_3}}\text{:=}0.1,A_{1_1,P_3}\left(\left(3.5\right]\right),{\sigma }_{A_{1_1,P_3}}\text{:=}0.3,A_{1_2,P_3}\left(\left(2,3\right]\right),{\sigma }_{A_{1_2,P_3}}\text{:=}0.6,A_{1_3,P_3}\left(\left[1,2\right]\right),{\sigma }_{A_{1_3,P_3}}\text{:=}0.8,\mathrm{and}A_{1_4,P_3}\left(\text{<}1\right),{\sigma }_{A_{1_4,P_3}}\text{:=}1.$

To accomplish this mapping the length of time the key has been in use is simply placed into the correct bucket and the corresponding score assigned. The cryptoperiod of a public key embedded in a certificate is, ignoring a pre-mature revocation, at least as long as, but not limited to the certificate's validity period, e.g., consider certificate renewals based on the same underlying key pair.

An interdependency exists between key length, the time the key has been in use (cryptoperiod) and support for Perfect Forward Secrecy (PFS). The longer the time has been in use, the more likely it is to be compromised. The longer the key length, the less likely it is to be compromised within a given time period. PFS helps ensure that compromise of a private key used in deriving session keys does not compromise previously derived session keys, thus helping to ensure long term confidentiality of the session even in the face of such a compromise. Support for PFS is represented by the key-exchange indicator in the negotiable cipher suites supported by a network service as mentioned in the description of property P₁ (TLS security). The key-exchange algorithm is encoded in the TLS cipher suite parameter (see IANA for a list of registered values) and indicated by KeyExchangeAlg in the normative description for cipher suites TLS_KeyExchangeAlg_WITH_EncryptionAlg_MessageAuthenticationAlg, see (TLS Configuration, Sec. 3.3 and Appendix B)

The combination of the security strength for the public key, the key's accumulated cryptoperiod and an optional support for PFS for an example embodiment is captured by the following table.

TABLE 1
Value lookup table for key strength, cryptoperiod and PFS support
	A10,P3	A11,P3	A12,P3	A13,P3	A14,P3
A00,P3	0	0	0	0	0
A01,P3   A02,P3   A03,P3   A04,P3	$\{\begin{array}{cc}\frac{{\sigma }_{A_{0_i,P_3}}+{\sigma }_{A_{1_j,P_3}}}{2},& 0\le i\le 4,0\le j\le 4,\mathrm{without}\mathrm{PFS}\\ & \mathrm{support}\\ {\sigma }_{A_{i,P_1}},& 0\le k\le 4,\mathrm{with}k\mathrm{according}\mathrm{to}\mathrm{the}\\ & \mathrm{PFS}\mathrm{security}\mathrm{strength}\end{array}\hspace{1em}$

With PFS being an embodiment of the cryptographic primitives the scoring of which is introduced for the property P₁ (above). PFS is scored according to the security strength bucket definitions for σ_AjP1 with 0≦j≦4

To complete the calculation of the overall score σ_P3 for property P₃, the hashing part of the certificate's signature algorithm (A_2,P3) can be scored (e.g., according to NIST's security strength assignment in Key Management Recommendations). Similarly to the key size evaluation, the score assignment can be given as:

$A_{2_0,P_3}\left(\text{<}80\right),{\sigma }_{A_{2_0,P_3}}\text{:=}0,A_{2_1,P_3}\left(80\right),{\sigma }_{A_{2_1,P_3}}\text{:=}0.4,A_{2_2,P_3}\left(112\right),{\sigma }_{A_{2_2,P_3}}\text{:=}0.6,A_{2_3,P_3}\left(128\right),{\sigma }_{A_{2_3,P_3}}\text{:=}0.8,A_{2_4,P_3}\left(192\right),{\sigma }_{A_{2_4,P_3}}\text{:=}0.9,\mathrm{and}A_{2_5,P_3}\left(256\right),{\sigma }_{A_{2_5,P_3}}\text{:=}1.$

Using a uniform weight for the attributes in the table, the attribute score σ_{A0×A1lookup,P3} can be obtained by a matrix lookup from Table 1, leading to a property score:

${\sigma }_{P_3}\text{:=}W_{A_0\times A_1\mathrm{lookup},P_3}·{\sigma }_{A_0\times A_1\mathrm{lookup},P_3}+W_{A_2,P_3}·{\sigma }_{A_2,P_3}\mathrm{where}\text{:}W_{A_0\times A_1\mathrm{lookup},P_3}\text{:=}0.8,W_{A_2,P_3}\text{:=}0.2\mathrm{and}{\sigma }_{A_2,P_3}\in \left\{{\sigma }_{A_{2_0,P_3}},{\mathrm{\dots \sigma }}_{A_{2_5,P_3}}\right\}$

from the paragraph above.

In one embodiment, the property P₄ (Revocation Infrastructure) might initially assign attribute scores based on the availability and accuracy of the revocation infrastructure employed by a certificate's issuer. The “better” the revocation infrastructure, the less likely it is that a revoked certificate will be determined to be unrevoked. In this context “better” can be defined by a relationship between Certificate Revocation List (CRL) Distribution Points (CDPs), see RFC 5280, Sec. 4.2.1.13, and Online Certificate Status Protocol (OCSP), see RFC 6960, responders assigned as revocation status access points for a specific certificate. The table below captures the attribute scores for a representative relationship.

TABLE 2
Value lookup for CDP vs. OSCP support
	CDP I	CDP II	CDP III	CDP IV	CDP V
OCSP I	1.0	0.9	0.7	0.2	$\{\begin{array}{cc}0.9,& \mathrm{if}\mathrm{subscriber}\\ 0,& \mathrm{if}\mathrm{subordinate}\mathrm{CA}\end{array}\hspace{1em}$
OCSP II	0.8	0.7	0.5	0.1	$\{\begin{array}{cc}0.6,& \mathrm{if}\mathrm{subscriber}\\ 0,& \mathrm{if}\mathrm{subordinate}\mathrm{CA}\end{array}\hspace{1em}$
OCSP III	0.6	0.5	0.3	0	$\{\begin{array}{cc}0.4,& \mathrm{if}\mathrm{subscriber}\\ 0,& \mathrm{if}\mathrm{subordinate}\mathrm{CA}\end{array}\hspace{1em}$
OCSP IV	0.3	0.2	0.1	0	0
OCSP V	0.7	0.5	0.4	0	0

• • Where:

CDP I: At least one CDP entry exists; the CRL (and optionally Delta-CRLs, if Delta CRLs comply with the applicable authoritative policy) retrieved from this CDP (or authoritative redirected CDPs, CDP redirection complies with the applicable authoritative policy) is valid (CRL is not expired, the CRL has been signed by an authorized entity, and the signature can be successfully validated); for subscriber certificates, the update interval (nextUpdate-thisUpdate) is less than or equal to seven days; for subordinate CA certificates the update interval is less than or equal to twelve months.

• • CDP II: At least one CDP entry exists; the CRL (and optionally Delta-CRLs, if Delta CRLs comply with the applicable authoritative policy) retrieved from this CDP (or authoritative redirected CDPs, CDP redirection complies with the applicable authoritative policy) is valid (CRL is not expired, the CRL has been signed by an authorized entity, and the signature can be successfully validated); for subscriber certificates, the update interval (nextUpdate-this Update) is greater than seven days, but less than or equal to ten days; for subordinate CA certificates the update interval is less than or equal to twelve months.
  • CDP III: At least one CDP entry exists; the CRL (and optionally Delta-CRLs, if Delta CRLs comply with the applicable authoritative policy) retrieved from this CDP (or authoritative redirected CDPs, CDP redirection complies with the applicable authoritative policy) is valid (CRL is not expired, the CRL has been signed by an authorized entity, and the signature can be successfully validated); either for subscriber certificates, the update interval (nextUpdate-thisUpdate) is greater than ten days or for subordinate CA certificates the update interval is greater than twelve months.
  • CDP IV: At least one CDP entry exists; the CRL (and optionally Delta-CRLs, if Delta CRLs comply with the applicable authoritative policy) cannot be retrieved from this CDP (or authoritative redirected CDPs, CDP redirection complies with the applicable authoritative policy) or is not valid (CRL is expired, the CRL has been signed by an unauthorized entity, or the signature cannot be successfully validated).
    • CDP V: A CDP entry does not exist.
  • And
  • OCSP I: At least one OCSP responder URL and/or a stapled response is provided (intercorrelation to A_1,P0 ((Multiple) Certificate Status Request)); the OCSP responder can be successful queried, the OCSP response is valid (OCSP response is syntactically correct, OCSP response has been signed by an authorized entity, and the signature can be successfully validated); for subscriber certificates, the update interval (nextUpdate-thisUpdate) is less than or equal to four days; for subordinate CA certificates the update interval is less than or equal to twelve months.
  • OCSP II: At least one OCSP responder URL is provided; the OCSP responder can be successful queried, the OCSP response is valid (OCSP response is syntactically correct, OCSP response has been signed by an authorized entity, and the signature can be successfully validated); for subscriber certificates, the update interval (nextUpdate-thisUpdate) is greater than four days, but less than or equal to ten days; for subordinate CA certificates the update interval is less than or equal to twelve months.
  • OCSP III: At least one OCSP responder URL is provided; the OCSP responder can be successful queried, the OCSP response is valid (OCSP response is syntactically correct, OCSP response has been signed by an authorized entity, and the signature can be successfully validated); either for subscriber certificates, the update interval (nextUpdate-thisUpdate) is greater than ten days or for subordinate CA certificates the update interval is greater than twelve months.
  • OCSP IV: At least one OCSP responder URL is provided; the OCSP responder cannot be queried or the OCSP response is not valid (OCSP response is syntactically incorrect, OCSP response has been signed by an authorized entity, or the signature cannot be successfully validated).
    • OCSP V: An OCSP responder URL is not provided.

The particular scoring uses policy guidelines applying to X.509 TLS server certificates, (e.g., see “Baseline Requirements for the Issuance and Management of Publicly-Trusted Certificates, v.1.3.0”, Forum Guideline, https://cabforum.org/baseline-requirements-documents, 2015, CA/Browser Forum, Sec. 4.9, 7.1.2.2, 7.1.2.3, “Guidelines For The Issuance And Management Of Extended Validation Certificates, v.1.5.5”, Forum Guideline, https://cabforum.org/extended-validation, 2015, CA/Browser Forum, Sec. 13) and then applies a heuristic assessment to arrive at the mapped scores. Assuming a uniform weight, the attribute score σ_P4 can be obtained by a matrix lookup from Table 2 above.

The initial property scores, P₀ . . . P₄ can then be combined using the weights given above according to the equation given above to get the initial security reliance score for this embodiment:

σ=0.2P₀+0.21P₁+0.15P₂+0.25P₃+0.2P₄

Surveys and Data Collection

After the initial scores have been calculated and stored in operation 204, operation 206 uses survey and data collection methods to gather information needed for calculating and updating both the values of attributes and/or properties and the scores related thereto. In addition, changed attributes and/or properties can be identified to add new or remove existing attributes and/or properties from consideration.

In one embodiment, information pertaining to the TLS landscape of an organization is inventoried by utilizing databases containing public network address associations for the organization, e.g., database of a network registrar, DNS server and reverse DNS databases, and WHOIS queries can be exploited to create a set of an organization's publicly visible network services. Examples of this are described below. If internal network services of an organization are targeted, access to the internal network is assigned to the information collecting system (FIGS. 3 and 4 discussed below). In this case, internal databases—e.g., internal DNS zones, IP address management databases—are queried to map out available services inside an organization.

As a result of connecting to these services, system, network protocol, and cryptographic configurations are explored, collected, aggregated, and stored for later analytics processing.

In a representative embodiment, configuration data is collected by attempting TLS handshakes. This allows for an evaluation of the TLS specific configuration similar to the property score evaluation of the previously described property P₀ (TLS Configuration) and P₁ (TLS Security). Then, by obtaining the certificates employed in securing the service, certificate specific security information is gathered similar to the evaluation of P₃ (Certificate Security). In addition the application protocol, e.g., HTTP over TLS (HTTPS), can be explored to gather further security specific application settings, e.g., HSTS enabling, public key pinning over HTTP (HPKP) similar to the evaluation of the property P₂ (Certificate Context) or a subset of attributes thereof

Turning to FIGS. 3 and 4 representative survey and data collection systems and methods will be described that are suitable for executing operation 206 of FIG. 2.

FIG. 3 depicts a representative architecture 300 to perform survey and data collection activities. In this representative architecture, a data collection and/or survey system 308 is connected to one or more systems (target systems 302, 304, 306) from which data is to be collected and/or surveys made. Connection can be made over a private network, public network, or combinations thereof as the type of connection doesn't matter as long as it is sufficient to allow the data collection/survey system 308 to collect the desired information. The data collection/survey system 308 interacts with the target systems 302, 304, 306 to identify cryptographic material and configuration information. The system operates as above, for example, to identify TLS information about the target systems. Thus, the data collection/survey system 308 can establish TLS connections with each system to identify all information needed. In some embodiments, multiple connections using multiple parameters are used to identify all of the configuration and cryptographic information that is desired. Thus, sufficient connection attempts can be made to identify the information used for analysis.

In other embodiments, in addition to or as an alternative to the above, information is collected from repositories, servers, or other systems/entities that may have already been collected. For example, application Ser. No. 14/131,635 entitled “System for Managing Cryptographic Keys and Trust Relationships in a Secure Shell (SSH) Environment,” assigned to the same assignee as the present application and incorporated herein by reference, identifies systems and methods for centralized management of cryptographic information such as keys and discusses a method of data collection from various systems in an SSH type environment. Such systems may have information that can be used to perform the requisite analysis and so can be a source of information.

As information is collected, the information can be categorized and stored for later evaluation as described above.

Turning next to FIG. 4, this figure illustrates an example deployment architecture 400, that sets a data collection/survey system (such as 308 of FIG. 3) into a cloud and/or service architecture. As illustrated in FIG. 4, the system is deployed in a cloud 402, which may be a private, government, hybrid, public, hosted, or any other type of cloud. Such a cloud deployment typically includes various compute clusters 412, 414, databases such as archival storage 418 and database storage 416, load balancers 404 and so forth. Such a cloud deployment can allow for scaling when multiple users/target systems 406, 408, 410 exceed capacity or when lesser capacity is needed to support the desired users/target systems 412, 408, 410. Furthermore, such an architecture can be used when the functionality provided by the system is offered as a service. Finally, the various users and/or target systems 406, 408, 410 are representative of the type of users and/or target systems that can utilize such a service. In the diagram target system 406 represents a single system, target systems 410 represent a small or moderate size deployment with multiple target systems either alone or tied together using some sort of network and target systems 408 represent a large scale deployment, possibly a cloud deployment or a company with multiple data centers, many servers, and/or so forth.

Returning now to FIG. 2, operation 206 represents collection of data and or conducting surveys of target systems (such as by the architectures in FIGS. 3 and/or 4) to gather information for analysis. Information gathered can include, but is not limited to:

• • IP Address
  • DNS name
  • X.509 Certificate(s) (End-entity and any issuing certificates provided)
  • SSL/TLS protocol versions supported
  • Cipher suites supported
  • HSTS support
  • Susceptibility to common vulnerabilities
  • Application type(s) (Web Server, Application Server, SSH server, etc.)

Adjustment of Scores and Learning Model Description

Once the information is collected in operation 206, operation 208 uses learning models, pattern recognition, statistical analysis and other methods to update attribute and/or properties values and scores based on various models. Specifically, operation 208 uses the information collected in operation 206 to calculate an update vector used in conjunction with an aggregation function to account for changes over time that should adjust attribute or other scores. The details of these processes are illustrated in FIG. 5.

As noted in conjunction with FIG. 1 above, attribute, property and overall scores can additionally be adjusted by applying statistical analysis, dynamic pattern recognition, and/or other learning algorithms and by evaluating additional context-sensitive data such as geographic location. One embodiment utilizes the principle that the security impact of a cryptographic primitive is related to its adoption rate relative to the baseline of growth of cryptographic material itself. Impact in this sense enhances the notion of security strength, which is based on a primitive's resilience against attacks. The following uses the hashing security primitive as an example of how a greater degree of adoption shows how the market trades off computational complexity with security impact.

In the Key Management Recommendations reference described above, the NIST identifies a security strength assignment of 256 bits for the signature hashing algorithm SHA-512, and a lower security strength of 128 bits for SHA-256. Both algorithms provide better security than SHA-1 (80 bits of security strength), but it is SHA-256 that has a higher adoption rate (due largely to the lack of support of public CAs for SHA-512). The higher adoption rate of SHA-256 over SHA-512 indicates that the additional increase in security strength for a single primitive like SHA-512 does not compensate for the additional computational complexity. The greater degree of adoption for a given primitive thus reflects its implementation impact.

Using the hashing algorithm of an X.509 TLS server certificate's signature as a representative example, the survey of publicly accessible network services secured by the TLS protocol provides the necessary data samples to assess adoption rate. In one example, A learning algorithm (see below) adjusts the initial attribute score assignment based on a hashing algorithm's security strength via its adoption rate according to a formula that captures the principle that low growth rates indicate either outdated (very) weak algorithms, or new and sparsely adopted ones, while high growth rates indicate (very) strong hashing algorithms. Assuming such survey was performed in 2012, the assigned values could be:

• • MD2: very weak algorithm; very low adoption rate
  • MD5: weak algorithm; low adoption rate
  • SHA-1: acceptable algorithm strength; high adoption rate
  • SHA-256: strong algorithm; low adoption rate
  • SHA-512: very strong algorithm; very low adoption rate

By continuously repeating the survey, the learning algorithm adjusts the hashing algorithm's attribute score assignment to reflect shifts in the hashing algorithm's growth rate and occasional updates to its security strength rating. The same evaluation in 2015, with the support of SHA-256 by public certification authorities (CAs) and introduction and approval of new algorithms, e.g., SHA-3 by NIST, might result in:

• • MD2: very weak algorithm; very low adoption rate
  • MD5: very weak algorithm; very low adoption rate
  • SHA-1: barely acceptable algorithm strength; medium adoption rate
  • SHA-256: strong algorithm; high adoption rate
  • SHA-512: strong algorithm; low adoption rate
  • SHA-3: very strong algorithm; initial security strength assignment

Assignments of attribute scores to a property and/or attribute can be automatically adjusted to reflect changes in the security landscape over time, as illustrated in process 500 of FIG. 5. The initial assignment of the attribute scores σⁱ can be updated to σⁿ in response to incoming information via the relationship:

σⁿ=ƒ(σⁱ,{right arrow over (Φ)})

where the (one or more dimensional) update vector {right arrow over (Φ)} is learned from incoming information, and ƒ is a function that aggregates the initial attribute score assignments and the update vector to produce a new attribute score.

As illustrated in FIG. 5, optional operation 504 can select an appropriate model for the adjustment vector {right arrow over (Φ)}. In one embodiment, the attribute score adjustment is made with an update vector {right arrow over (Φ)} that assigns a value in the interval [0,1] to doubling times (how long it takes for the population with a particular feature to double in size) derived from an exponential model of the growth of a specified subset of certificates over time (see FIG. 6), and an aggregating function ƒ taken to be the geometric mean. In this embodiment, {right arrow over (Φ)} compares the doubling time of a subset of certificates (t_subset) to the doubling time of all certificates (t_{all_certificates}), and assigns a value between 0 and 0.5 to certificate subsets with a doubling time longer than the overall certificate doubling time, and a value between 0.5 and 1 to certificates with a doubling time shorter than it:

$\overrightarrow{\Phi }\left(t_{\mathrm{subset}}\right)=2^{-\left(\frac{t_{\mathrm{subset}}}{t_{{\mathrm{all}}_{—}\mathrm{certificate}}}\right)}$

and the aggregating function ƒ is the geometric mean defined:

$f\left(X_1,X_2,\dots ,X_n\right)={\left(\prod _{i=1}^nX_i\right)}^{1\text{/}n}$

FIG. 6 illustrates the update vector function {right arrow over (Φ)}.

In one embodiment, the attribute score adjustment is calculated for A_2,P3 the hashing part of the certificate's signature algorithm for property P₃. The initial property score is assigned to a certificate based on the NIST security strength assignment of its hashing algorithm as described above. The property score is then updated in response to updated information that reflects changes in the impact the algorithm is having on the community, as quantified by the algorithm adoption rate. This adoption rate is learned from periodic large-scale scans of certificates (e.g., operation 206, FIG. 2). An exponential model is fitted to the cumulative number of certificates employing a particular hashing algorithm as a function of certificate validity start date. The exponent of the model yields a measure of the algorithm adoption rate (operation 506). This adoption rate may then be used in the function {right arrow over (Φ)} to calculate the update vector (operation 508). The update vector is then combined with the initial value to calculate the new score (operation 510). For example, we may observe that in 2015, the number of hashing algorithms with a NIST security strength assignment of 128 is doubling every 2 years

$\left(t_{A_{2_3}}=2\right),$

the number of algorithms given a strength of 256 is doubling every 4 years

$\left(t_{A_{2_5}}=4\right),$

and the time taken for the total number of certificates to double is 5 years (t_{all_certificates}=5). The initial values of the attribute scores for certificates with hashing algorithms assigned NIST security scores of 128 and 256 would be updated in response to the empirical doubling times via (operation 508):

${\overrightarrow{\Phi }}_{A_{2_3,P_3}}=2^{-\left(\frac{t_{A_{2_3}}}{t_{{\mathrm{all}}_{—}\mathrm{certificates}}}\right)}=2^{-\left(\frac{2\mathrm{years}}{5\mathrm{years}}\right)}\cong 0.76,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}=2^{-\left(\frac{t_{A_{2_5}}}{t_{{\mathrm{all}}_{—}\mathrm{certificates}}}\right)}=2^{-\left(\frac{4\mathrm{years}}{5\mathrm{years}}\right)}\cong 0.57,$

so that for the NIST 128 algorithms (operation 510):

${\sigma }_{A_{2_5,p_3}}^n=F\left({\sigma }_{A_{2_5,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}\right)={\mathrm{geometric}}_{—}\mathrm{mean}\left({\sigma }_{A_{2_3,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_3,P_3}}\right)={\left(\Pi \left({\sigma }_{A_{2_5,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_3,P_3}}\right)\right)}^{1\text{/}2}={\left(0.8·0.76\right)}^{1\text{/}2}\cong 0.78$

and for the NIST 256 algorithms (operation 510):

${\sigma }_{A_{2_5,P_3}}^n=f\left({\sigma }_{A_{2_5,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}\right)={\mathrm{geometric}}_{—}\mathrm{mean}\left({\sigma }_{A_{2_5,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}\right)={\left(\Pi \left({\sigma }_{A_{2_5,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}\right)\right)}^{1\text{/}2}={\left(1·0.57\right)}^{1\text{/}2}\cong 0.75$

In this example the algorithms with a NIST security strength assignment of 256, while given an initial score greater than the NIST 128 algorithms, are nevertheless given a lower final score than the NIST 128 algorithms because of their slower adoption rate and lower impact on the community in 2015.

We could potentially see a reversal of this evaluation in 2020 if we observe that the number of hashing algorithms with a NIST security strength assignment of 128 is doubling every 4 years

$\left(t_{A_{2_5}}=4\right),$

the NIST 256 algorithms are doubling every 1.5 years

$\left(t_{A_{2_5}}=1.5\right),$

and the total number of certificates is doubling every 4.5 years (t_{all_certificates}=4.5). The update vectors would be (operation 508):

${\overrightarrow{\Phi }}_{A_{2_3,P_3}}=2^{-\left(\frac{t_{A_{2_3}}}{t_{{\mathrm{all}}_{—}\mathrm{certificates}}}\right)}=2^{-\left(\frac{4\mathrm{years}}{4.5\mathrm{years}}\right)}\cong 0.54,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}=2^{-\left(\frac{t_{A_{2_5}}}{t_{{\mathrm{all}}_{—}\mathrm{certificates}}}\right)}=2^{-\left(\frac{1.5\mathrm{years}}{4.5\mathrm{years}}\right)}\cong 0.79,$

so that for the NIST 128 algorithms (operation 510):

${\sigma }_{A_{2_3,P_3}}^n=f\left({\sigma }_{A_{2_3,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_3,P_3}}\right)={\mathrm{geometric}}_{—}\mathrm{mean}\left({\sigma }_{A_{2_3,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_3,P_3}}\right)={\left(\Pi \left({\sigma }_{A_{2_3,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_3,P_3}}\right)\right)}^{1\text{/}2}={\left(0.8·0.54\right)}^{1\text{/}2}\cong 0.66$

and for the NIST 256 algorithms (operation 510):

${\sigma }_{A_{2_5,P_3}}^n=f\left({\sigma }_{A_{2_5,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}\right)={\mathrm{geometric}}_{—}\mathrm{mean}\left({\sigma }_{A_{2_5,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}\right)={\left(\Pi \left({\sigma }_{A_{2_5,P_3}}^i,{\overrightarrow{\Phi }}_{A_{2_5,P_3}}\right)\right)}^{1\text{/}2}={\left(1·0.79\right)}^{1\text{/}2}\cong 0.89$

The NIST 256 algorithms are now given a much higher score than the NIST 128 algorithms; a reflection of both the faster adoption rate and the higher initial value of the attribute score for the NIST 256 algorithms. In general, this approach can be applied to any attribute score associated with a property of certificates that may improve or be updated over time.

In this example, a particular update function was identified to adjust a parameter that conforms well, within a fixed time window, to an exponential model. Different models may be used to adjust other properties and/or attributes over time that are better described with a non-exponential model, resulting in selection of a different model as part of operation 504.

If update vector identified in operation 208 would result in updated scores, then the “Yes” branch is taken out of operation 210 and the scores are recalculated in operation 212. Operation 212 is performed according to the discussion around setting the initial scores as disclosed above. In other words, the scores for various attributes are calculated and combined according to the functions disclosed above to yield property scores for each property. The property scores are then aggregated according to the weighted sum disclosed above to yield an overall score. If further aggregation is desired (across a system, cluster of systems, cryptographic material holder, subsidiary, company, etc.), then the further aggregation is performed.

Statistical Sampling and Geographic/Contextual Adjustments

The overall score σ, calculated as described in the previous paragraphs, can in addition be further affected by a statistical analysis, by applying dynamic pattern recognition and by evaluating additional context-sensitive data. In one embodiment, statistical anomaly probing is part of operation 208 (illustrated as process 502 of FIG. 5) and examines the likelihood of the specific occurrence of the cryptographic material and/or the likelihood of specific context configuration for the cryptographic material when compared to a test group of similar samples.

Operation 512 of FIG. 5 selects the context-sensitive factors and attributes that will be used to calculate the security anomaly score. In one embodiment the geo-location context of a collected X.509 TLS server certificate might be evaluated as part of the anomaly probing. The following example helps explain how this arises and the impact it can have. Different national regulatory bodies recommend the use of otherwise less commonly applied cryptographic primitives, e.g., the Russian GOST specifications R. 34.10, 34.11, etc. For application of the GOST specifications in X.509 certificates see RFC 4491. Which regulatory body applies often depends on the geo-location context of the certificate. Using the GOST specifications as a representative example, in one embodiment, X.509 TLS server certificates whose signature has been produced with such a GOST-algorithm might be further examined in regards to the certificate's ownership—specifically the country code part of the certificate's subject distinguished name—and IP address provenience, i.e., the geo-location metadata for the IP address for which the certificate has been employed.

Given a 2×2 contingency table counting the number of certificates that do or do not use a GOST signature algorithm, and that are located inside or outside of Russia, we can assign an anomaly score Ω to a certificate that reflects the interaction between the use of the GOST signature algorithm and the certificate's geo-location. For example, in a collection of observed certificates the abundances of the possible combinations of these properties (relative to the total number of certificates) may be as given in the table below:

TABLE 3
Example X.509 Certificates Using
GOST Signature Algorithms
		Inside	Outside of
		Russia	Russia
	Uses GOST signature algorithm	0.02	0.005
	Does not use GOST signature	0.05	0.925
	algorithm

which we write:

$M=\left(\begin{array}{cc}0.02& 0.005\\ 0.05& 0.925\end{array}\right)$

The anomaly score for a certificate that uses the GOST signature algorithm, and is found outside of Russia, would be calculated on the basis of the conditional probability that the signature algorithm is “GOST” given that the geographic region is not Russia (operation 514). This probability is given by:

$p=p\left(\mathrm{GOST}|{\mathrm{Outside}}_{—}\mathrm{Russia}\right)=\frac{M_{1,2}}{\sum _{i=1}^2M_{i,2}}=\frac{0.005}{0.005+0.925}\cong 0.0054$

In embodiments disclosed herein, the anomaly score is selected to remain near 1 except in the case of a very anomalous certificate. In other words, applying this approach small values of the conditional probability described above identify anomalous certificates, but differences between large and middling values of this probability are unlikely to indicate a meaningful difference between certificates. For this reason in one embodiment the anomaly score is calculated (operation 516) from the conditional probability via a sigmoidal function that exaggerates differences between low conditional probabilities, but is largely insensitive to differences between probabilities in the mid and high range:

$\Omega \left(p\right)=\frac{1-e^{-\mathrm{sp}}}{1+e^{-\mathrm{sp}}}$

where s is parameter that controls the range of probabilities to which Ω is sensitive. In a representative embodiment, a suitable value for s would be 100, chosen to tune the range of probabilities to which the anomaly scoring function is sensitive. FIG. 7 plots Ω(p) for s=100. Using this function, the anomaly score for a certificate found using the GOST signature algorithm outside of Russia (the p(GOST|Outside_Russia)=0.0054 from above) would be given by (operation 516):

$\Omega \left(p\left(\mathrm{GOST}|{\mathrm{Outside}}_{—}\mathrm{Russia}\right)\right)=\frac{1-e^{-{100}^{*}0.0054}}{1+e^{-{100}^{*}0.0054}}\cong 0.26$

On the other hand, for a GOST certificate that was found in Russia, Ω would be given by (operations 514 and 516):

$p=p\left(\mathrm{GOST}|{\mathrm{Inside}}_{—}\mathrm{Russia}\right)=\frac{M_{1,1}}{\sum _{i=1}^2M_{i,1}}=\frac{0.02}{0.02+0.05}\cong 0.0286$ $\Omega \left(p\left(\mathrm{GOST}|{\mathrm{Inside}}_{—}\mathrm{Russia}\right)\right)=\frac{1-e^{-{100}^{*}0.286}}{1+e^{-{100}^{*}0.286}}\cong 1$

Thus Ω assigns a score very close to 1 to the certificate with the unsurprising location within Russia, but gives a significantly smaller value to the anomalous certificate that uses the GOST signature algorithm outside of Russia.

The anomaly function, the initial security reliance score, and debasing constant Δ, if any of the debasing conditions are met, are used to determine an adjusted security reliance score through the equation at the beginning of the disclosure:

$\sigma \text{:=}\{\begin{array}{cc}\Delta ,& \mathrm{if}\mathrm{debasing}\mathrm{condition}\mathrm{is}\mathrm{met}\\ \Psi \left\{\sum _{i=0}^n{\sigma }_{P_i}·W_{P_i},\Omega \right),& \mathrm{otherwise}\end{array}$

As explained above, the mapping function, Ψ, combines the security reliance score, and the anomaly score to adjust the security reliance score for the information contained in the anomaly score. In one embodiment, the function, Ψ, selects the minimum between the security reliance score and the anomaly score. Thus:

$\Psi \left(\sum _{i=0}^n{\sigma }_{P_i}·W_{P_i},\Omega \right)=\min \left(\sum _{i=0}^n{\sigma }_{P_i}·W_{P_i},\Omega \right)$

In another embodiment, the function, Ψ, calculates the mean of its inputs. Thus:

$\Psi \left(\sum _{i=0}^n{\sigma }_{P_i}·W_{P_i},\Omega \right)=\frac{1}{2}\left(\left(\sum _{i=0}^n{\sigma }_{P_i}·W_{P_i}\right)+\Omega \right)$

Once the value for any Ω is calculated (operation 502), the “yes” branch out of operation 210 is also triggered and the scores recalculated in operation 212 and stored in operation 214 as previously described.

If no changes are detected as part of operation 210, the “no” branch is taken and the system can wait until new information is collected that could impact the scores.

Other Uses for Survey Data

The information collected as part of survey data collection operation 206 can also be used for other (optional) purposes such as generate survey reports (operation 216 discussed below) and identifying new attributes/properties that should be included as part of the scoring system (operation 218).

Identification of new attributes/properties can occur based on analysis of the collected data (operation 206). For example, the ongoing data collection may discover an X.509 TLS server certificate that employs a new and previously unseen signature algorithm. In one embodiment, the attribute score programmatically associated with the new signature algorithm would be set to a default value of 0.5. In subsequent data collections, it would become possible to estimate the adoption rate and doubling time for the new algorithm. If the new algorithm begins to be highly adopted, this will be reflected in the update vector and lead to the adjustment of the corresponding attribute score toward a higher value in indication of the high security impact the algorithm is having. If, on the other hand, the algorithm does not gain widespread adoption, the corresponding attribute score will drop in reflection of the low impact of the new signature algorithm.

Once new attributes and/or properties have been identified as part of operation 218, the “yes” branch is taken out of operation 220 and initial values for the attributes are set and the initial scores calculated. As indicated previously, in some embodiments, attribute scores for particular properties are calculated in different ways (i.e., using different functions) for different properties (e.g., not every embodiment uses the same functions to aggregate property scores for all properties). Examples of these functions have been discussed above. If the system identifies new attribute(s), functionality to handle the new attribute(s) can be added to the system to calculate the new scores/property scores if desired. Periodically, properties are re-defined and/or created by aggregating different existing and/or new attributes. Likewise, new implementations of cryptographic primitives are integrated into the corresponding security property's attribute by a manual initial security strength assignment, e.g., NIST's finalization of the cryptographic hashing standard SHA-3.

Although operation 218 and operation 220 are specified in terms of “new” attributes and/or properties, some embodiments also identify whether existing attributes should be removed. Additionally, or alternatively, attributes that no longer apply can be debased using debasing conditions, as previously described above.

Use of the Security Reliance Score

The security reliance score, or a subset of its property or attribute scores in a variety of particular combinations, can be aggregated and further customized to target the specific landscape of an organization, such as depicted as part of operation 216 and as described above (e.g., further aggregation of the security reliance scores).

Many organizations lack the ability to identify even the most egregious cryptographic key-related vulnerabilities that need to be addressed. Evaluation is accomplished in some embodiments by calculating a security reliance score, as indicated above. The calculated scores allow for an ordering by worst configurations encountered for the network services provided by an organization or partitions of it.

FIG. 8 illustrates how the security reliance score, or aggregated security reliance scores (i.e., aggregated across a system, business line, enterprise and/or business vertical) can be used to calculate a representative vulnerability scale. In this discussion below, security reliance score will be used although it is understood that the same disclosure applies equally to aggregated security reliance scores. Such a vulnerability scale can be derived from a security reliance score by placing the scores on a relative continuum, and setting thresholds for the various “levels” of vulnerability in order to “bucketize” a particular security reliance score into a particular vulnerability level. Additionally, or alternatively, specific causes may call for a particular place on the vulnerability scale. Thus, examining the attribute, property and overall scores and identifying the aspects that are causing an attribute score may give rise to a particular placement. For example, if the P₀ (TLS configuration) score described above is particularly low, an examination may reveal that the reason is that attribute A_2,C0 (Renegotiation) as the TLS Insecure Renegotiation enabled (thus giving it a score of only 0.3). This factor can then be identified as a cause of the low score.

Such an examination also yields suggestions on how to improve the scores and can further identify changes that will have the biggest impact. Thus, the examination may yield information that can be presented to a system administrator, or other user of the system, to help them diagnose and correct security issues.

The representative vulnerability scale in FIG. 8 has six categories, indicating increasing levels of vulnerability. These can be presented in various ways including having symbols (such as those illustrated as part of levels 800, 802, 804, 806, 808, and 810) and/or color coding to visually convey a sense of urgency associated with increasing levels of vulnerability. The various illustrated levels include:

• • 1. Secure 800: The entities' certificate configuration and network service configuration are secure.
  • 2. At Risk 802: The entities' certificate configuration or network service configuration does not follow security best practices and places the organization at risk of being exploited.
  • 3. Vulnerable 804: The entities' certificate configuration or network service configuration is vulnerable to known exploits.
  • 4. Critical 806: The entities' certificate configuration or network service configuration is vulnerable to common exploits.
  • 5. Hazardous 808: The entities' certificate configuration or network service configuration is vulnerable to several common exploits.
  • 6. Exposed 810: The entities' certificate configuration or network service configuration is exposed to several, severe common exploits, action should be taken immediately.

Some embodiments comprise a model ‘calculator’ or ‘evaluator’ that dynamically highlights how specific TLS configuration settings can improve or decrease ones overall TLS security posture. Such an interactive tool can utilize stored security reliance scores (overall, property, attribute, aggregated, and so forth) to allow a user to interactively evaluate and investigate scores (at various levels), aggregate and drill into scores and their components, evaluate underlying causes for the various security reliance scores and associated vulnerability levels, and investigate various configurations.

By presenting an interactive tool, that allows trying out different configuration settings, a customer is enabled to decide how to increase his overall security rating by focusing on settings with the biggest impact.

In addition to an interactive tool, embodiments may automatically recommend settings that, if changed, will have an impact on the overall security rating. Such recommendations can be based, for example, on the analysis above (e.g., identifying settings that have the biggest contribution toward an attribute score and then identifying which values that, if changed, will have the biggest impact on an attribute score).

Security scoring results for organizations, as described above, can be further grouped and aggregated by standard industry hierarchies, e.g., MSCI's Global Industry Classification Standard. Such a scoring aggregation can allow entities to compare their achieved security score with peers in the same industry area.

Furthermore, security reliance scores can be used in a recommendation system, as discussed below.

Recommendation System

As discussed above, security reliance scores are associated with cryptographic material. These security reliance scores can be at different levels and aggregated across levels all the way to an ultimate aggregated score. For example, all the cryptographic scores for the cryptographic material in an entire company can be aggregated to arrive at a security reliance score for the company. Examples of different levels that can be aggregated include, but are not limited to, an industry, company, subsidiary, location, facility, system, group(s) of systems, and so forth. In this context aggregation means some form of many-to-one mapping such as a weighted sum, average, median, and so forth. Many different ways of aggregating scores are known to those of skill in the art and any such mechanisms can be used.

For purposes of description, the security reliance scores will be identified as being stored in a security reliance score database. This database may be a single database or may be multiple databases with different sets of security reliance scores stored in different databases. The actual implementation does not matter as long as the system has access to the appropriate information. Representative examples of a security reliance score database can be found in data store 1412 or 1414 of FIG. 14 or 416 or 418 of FIG. 4.

In the context of a recommendation system, some cross section of cryptographic material of interest and their attendant scores can be evaluated in order to help administrators, users, or others that have responsibility for a cross section of the cryptographic material improve the associated scores and thus improve the security reliance, trustworthiness, and/or reliability of the underlying cryptographic key material, protocols, systems, process configurations and so forth.

Security reliance scores of interest (i.e., some cross section of scores that are to be evaluated) are retrieved and aggregated. Statistics and/or metrics associated with the scores can be presented to a user, administrator, or other individual who has authority to evaluate the scores. In one example, the scores falling into designated percentile ranges are counted and the percentage of scores in each range is calculated. The aggregated scores along with the overall aggregated scores at different levels can be presented, if desired. This allows an individual exploring an aggregate score to evaluate various combinations and identify specific targets for improvement. These security reliance scores can be compared to a comparison set to give the user, administrator, etc. a baseline for improvement. Improvement can be associated with one or more ‘improvement goals’ such as increasing the percentage of scores that are “good,” e.g., expressed as a range of percentiles with desirable scores, reducing the percentage of scores that are “bad,” e.g., expressed as a range of percentiles with undesirable scores, improving the aggregate score at a particular level, or other goals that would improve the security reliance scores in some fashion. In non-interactive embodiments, the system can monitor events and automatically implement actions consistent with the improvement goals.

FIG. 9 illustrates an example user interface 900 for guiding a user through security reliance score improvement on a selection of cryptographic key entities. This user interface includes a region 902 that allows the user to select a sample (sub)set of the security reliance database, as a basis for comparison. This selection of comparison material is referred to as the set of comparison cryptographic key material. The individual items in 904 e.g., reflect the security reliance database's full comparison set, “Full comparison set”, and subsets of it. “Comparison subset 1” may in one embodiment be the subset defined by organizations belonging to the same vertical as the user's, and “Comparison subset 2” may be the subset restricted to organizations in the same geographical region as the user's and so forth. The individual items 904 also show a comparison set of the user's cryptographic key material or a subset thereof. The disclosure is not limited in this manner and the comparison set of data (i.e., items selected in region 902) can be any set or subset that is desired. The individual items 904 are presented in such a way that the user is able to select one or more entries. This can be with radio buttons, check boxes that include/exclude different items, queries, filters, and so forth.

Region 906 allows the user to select a set of user cryptographic key material that will be considered for comparison to the set of comparison cryptographic material and for improvement. As shown in FIG. 9, such selection can be through various mechanisms. In some embodiments a user can enter one or more filter expressions, e.g., as provided by database query expressions like the standard query language (SQL) as shown by the filter entry region 908. Additionally, an area 910 can be provided that allows a user to select particular cryptographic key material (i.e., sets, subsets or individual cryptographic key material) for inclusion/exclusion. The filter(s) 908 and selection(s) 910 can work together such as allowing a user to enter a filter expression to select a set of cryptographic material and then select/deselect individual cryptographic material within the set retrieved by the filter/query to identify the set of user cryptographic key material for comparison and improvement. Additionally or alternatively, filters can be represented and/or entered graphically instead of requiring entry of a query, such as by using any of the various techniques that are known to those of skill in the art that help users build queries or filter data sets.

As the comparison set of cryptographic key material and/or user set of cryptographic key material are selected, one or more metrics that describe the set(s) can be presented to the user to give the user information on the scores of the set(s). In one example embodiment, with every addition to the set(s) of selected entries one or more a panels with statistics on the selection so far is updated. In FIG. 9, the statistics are presented in panel 912 and panel 914. In the example of FIG. 9, the panel 912 presents the proportion of the set of user cryptographic key material in defined percentile ranges of the security reliance overall score. For example, various ranges can be defined, selected, or otherwise specified by the user and/or system and the percentage (or number or some other aggregation) of the security reliance scores of the selected group(s) falling into each range can be displayed. The percentile ranges can be derived, for example, from the comparison set and the actual percentages of the user set in the percentile ranges can be displayed. In the illustrated embodiment 13% of the selected user cryptographic key material scores fall into percentile 1 (say the interval [0^th-10^th] of the comparison set), 80% of the selected user cryptographic key material scores fall into percentile 2 (say the interval (10^th-30^th] of the comparison set), and 7% of the user cryptographic key material scores fall into percentile 3 (say the interval (30^th-50^th] of the comparison set). This is all by way of example, and other statistics can also be displayed such as comparison statistics for another cross-section of scores (such as how the selection stacks up against the remainder of the non-selected scores, an entire enterprise, industry, department, or other cross-section such as the set of comparison of cryptographic key material), or any other information that would be useful in helping the user understand the security reliance scores of the selected cross-section.

In the example of FIG. 9, panel 914 contains averages for selected sets. Thus, panel 914 displays the average overall score for the comparison set of cryptographic key material 916, which is illustrated as 0.8, the average overall score for all cryptographic key material the user is responsible for 918, which is illustrated as 0.6, and an overall average of the user selected, i.e., the set of keys selected in 906, cryptographic key material 920 for, which is illustrated as 0.4. While averages are used as representative examples, other statistics such as a median or other aggregation can be used in lieu of or in addition to averages. Additionally, or alternatively, metrics can be shown for other sets/subsets of cryptographic key material.

Calculation of the displayed statistics is well within the knowledge of those of skill in the art once the relevant set of scores for which statistics are to be calculated and displayed. For example, in panel 912, the percentage (or number) of scores in each percentile range is calculated by counting the number of scores in the relevant set in each percentile range and then, if a percentage is desired, dividing by the total number of scores in the set and multiplying by 100. Similarly, an average, median, minimum, maximum, or any other similar metrics that are known can be calculated and displayed, such as in panel 914, to allow the user to assess information about a relevant set of scores. Comparison of any such metrics between the comparison (sample) set of scores and the user set of scores will allow a user to assess relative security strength of the user scores vs. the comparison set, as described herein.

Once the user has selected the set of user cryptographic key material (i.e., from panel 906) and the set of comparison cryptographic key material (i.e., from panel 902), the system can perform various methods and calculations to recommend actions that will improve the security reliance scores and the resulting statistics based on a set of improvement metrics. Primary improvement metrics may be increasing the average security reliance overall score of the selection of cryptographic material, increasing the proportion of the selection of cryptographic material in the top percentile range of the security reliance overall score, decreasing the proportion of the selection of cryptographic material in the lowest percentile range of the security reliance overall score, decreasing some sort of dispersion metric like the variance, increasing or decreasing some other metric, combinations thereof or some other appropriate objective. One or more user selected primary improvement metrics are used in performing calculations and making recommendations to the user. In FIG. 9, the primary improvement metric(s) are selected in panel 922. Example primary improvement metrics include increasing the number/percentage of cryptographic material in a particular percentile range, decreasing the number/percentage of cryptographic material in a particular percentile range, improvement of a particular metric like average score, decreasing some metric like a variance measure, and combinations thereof

In addition to identification of the primary metric(s) that will help improve the security reliance scores (i.e., selected in panel 922), the user can opt for a secondary improvement metric for which an optimization can be performed as explained below. In panel 924, the system displays secondary metrics that can be used in conjunction with the primary metrics in performing calculations and making recommendations to the user. In some instances, selection of a primary metric in panel 922 may trigger a change in the secondary metrics available for selection in panel 924. In other words, depending in some instances and in some embodiments, not all combinations of primary and secondary metrics may be useful in performing calculations and making recommendations. In many instances, the secondary metric(s) can represent an additional constraint in the improvement goal, as explained further below. Example secondary improvement metrics include minimizing cost, maximizing a metric like average score, matching the most common attribute(s), and combinations thereof. In this sense, minimizing and maximizing may not be a global minimum or maximum, but rather a choice that, when compared to other choices, lowers or increases the corresponding secondary metric like cost, average score, variance or other secondary metric, while accomplishing the primary improvement metric. A secondary doesn't always need to be selected in all embodiments.

The user's “improvement goal” comprises the primary improvement metric(s) taken together with the selected secondary metric(s), if any. As noted above, the secondary metric(s) often represent a measurable constraint. This constraint is applied in order to resolve the preference of attribute choice for the exemplary model. For example, a user's improvement goal may consist of the improvement metric “improving the overall average score” for the selected user cryptographic keys, and the secondary metric “minimize associated costs”. Alternatively, the improvement goal could consist of the improvement metric “increasing the proportion of the selection of cryptographic material in the top percentile range” with “maximize average overall score” as a secondary metric.

For each improvement goal of interest to the user, one or more recommended actions reflect the result of a computed improvement potential. In FIG. 9 panels 926, 928, 930 and 932 display the resulting impacts, labeled “Primary improvement impact X” and “Secondary improvement impact X” (if applicable) in each of the panels. The improvement impacts displayed in the respective panels represent the respective improvement potential associated with applying one of four different actions, “Action 1”, “Action 2”, “Action 3”, and “Action 4”, as displayed in the respective panel. The primary and secondary improvement impact for a particular panel is derived from the resulting exemplary model if the indicated action is taken. For example, let the primary metric be decrease the proportion of the selection of the user's TLS server certificates in the lowest percentile range of the security reliance overall score, and the secondary metric be maximize the average security reliance overall score, the, Action 1 may be the recommendation to replace domain vetted (DV) certificates by extended validation (EV) certificates, Action 2 may be the recommendation to reconfigure the servers employing the corresponding certificates, Action 3 may be the recommendation to extend the DNS resource records associated with the host and or domain names of the corresponding certificates, and Action 4 may be the recommendation to patch or upgrade a security library used by the servers who employ the corresponding certificates. The number of actions displayed and their impacts can vary according to the primary and secondary metric(s) selected.

The system can provide an interface element that will allow the user to see the impact of one or more selected actions. The primary and secondary impacts (if applicable) as displayed in panels 926, 928, 930 and 932 can be any indication that allows the user to assess the impact of the recommended action. For example, if the improvement goal comprises a primary metric of decreasing the number of certificates with a score in the lowest percentile and a secondary metric of improving the overall score of all user certificates, the primary impact and/or secondary impact may comprise metrics that show how many certificates are moved out of the lowest percentile and the secondary impact may be how much the overall score is increased. Similarly, rather than absolute values (i.e., the number of certificates in the lowest percentile and the overall score), some metric of relative change can be displayed, such as percentage improvement/decrease, absolute improvement/decrease, and so forth. Combinations of more than one such metric can also be displayed for the primary and/or secondary impact.

The system can also display costs associated with a particular action. Thus in FIG. 9, panels 926, 928, 930, and 932 also display an “estimated additional cost” field. This field can be calculated by aggregating the costs associated with the recommended action. As explained below, costs can either be a monetary cost or some other cost such as complexity/ease of implementation, time to implement, and so forth, or a combination of both.

If a user decides on one or more courses of action, the user can activate an appropriate user interface element to trigger at least one process aiming at accomplishing one or more of the recommended actions. In FIG. 9 such interface elements are represented by the “Apply” buttons in panels 926, 928, 930 and 932. Such an action can, for example, kickoff a workflow, invoke a Security Information & Event Management (SIEM) process, script, revoke and rotate a key, install a patch, redirect network traffic, reset a server's system environment, start/restart/shutdown a service, or any other action that is aimed at accomplishing one or more of the selected recommended actions.

FIG. 10 illustrates a suitable method 1000 for calculating the improvement potential (also referred to as improvement impact as displayed in FIG. 9) for a selected cross-section of security reliance scores. The method begins at operation 1002 where the system obtains the user cryptographic key material and comparison cryptographic key material. In some embodiments, this occurs as described in conjunction with FIG. 9 above, with the system receiving user selections of which underlying of cryptographic material, protocols, systems, process configurations, and/or other entities, along with their security reliance scores should be included in the two sets of key material. The user and comparison sets of cryptographic key material may also be obtained from some other sources such as being associated with an automated running of the process such as through a triggering event, a batch process, or in some other manner. Automated use of the process illustrated in FIG. 10 is discussed in greater detail below.

In operation 1004 the system calculates and/or displays statistics and/or metrics associated with the selected cross-section. If the process is being run in a fashion that allows display of the calculated statistics (i.e., such as in an interactive manner, or in a process where information is displayed/printed), the calculated statistics may then displayed as described in conjunction with FIG. 9 above. The actual calculation of the statistics was described above where the various scores are calculated and can be aggregated at various levels.

Operations 1002 and 1004 can be repeated as necessary if the system is being used in an interactive manner where the user adjusts selections, for example, through a user interface. Alternatively, the system can perform operations 1002 and 1004 as part of a process that does not require user interaction. In such an embodiment, the cross section of scores can be retrieved from an input file or input by some other process or system. Such operation is described further below. In this situation, it may not be necessary or advisable to display the statistics/metrics.

Operation 1006 creates an exemplary model so that the improvement potential for particular cryptographic key material can be calculated. Once a specific improvement goal, i.e., a primary and secondary improvement metric (if any), and the user and comparison data sets obtained, the attributes of the exemplary model are calculated from the attributes of the key material in that cross-section of the security reliance database (e.g., data store 1412 or 1414 of FIG. 14 or 416 or 418 of FIG. 4).

Turning for a moment to FIG. 11, a method 1100 for creating an exemplary cryptographic key material model will be described. In FIG. 11, 1102 illustrates a notional representation of metadata associated with cryptographic key material. For example, there may be some sort of optional identifier, a set of attributes, a score and other metadata associated with the cryptographic key material. In this illustration, for creation of the exemplary model, the cryptographic key material will have an ID, a set of attributes and a score, although the ID is used only to help illustrate what happens to various attribute sets in the method.

There are two basic operations in creating an exemplary cryptographic key material model, which are labeled as 1104 and 1112 in FIG. 11. The first operation is to select a target comparison set from the comparison set of cryptographic key material. The target comparison set is a subset of the comparison set of cryptographic key material that will be used as the basis for the model. This subset is representative of the desired objective under the primary improvement metric of the improvement goal and is called the target comparison set. The target comparison set represents the subset of comparison cryptographic key material which will be examined for attributes to create the exemplary model and is typically selected based on desired attributes, given the primary improvement metric of the improvement goal.

Operation 1104 illustrates selecting a target comparison set from the comparison set. How the target comparison set is selected depends on the primary improvement metric and is generally the subset the administrator is desiring to move things into. For example, if the primary improvement metric is to move scores into a designated percentile, the target comparison set is the subset of comparison scores in that percentile. If the primary improvement metric is to move scores out of a designated percentile, the target comparison set is everything but that percentile. If the primary improvement metric is to increase a metric, the target comparison set consists of all comparison key material with values for that metric above the appropriate cut-off. As an example, if the primary metric is to increase the average score, the target comparison set consists of all comparison key material with values for the security reliance score above the average security reliance score of the set of user key material. If the primary improvement metric is to decrease a metric, then the target comparison set consists of all comparison key material with values for that metric below the appropriate cut-off. As an example, if the goal is to reduce a dispersion metric such as a measure of variance within the various cryptographic attributes, the target comparison set would be the set of comparison key material whose attributes could result in a variance that is lower than the desired dispersion metric.

For purposes of illustration, assume that the primary improvement metric is to increase the average score. Thus, to select the target comparison set, the comparison set is ordered by score and those items with scores above the average of the reliance score of the set of user keys are selected for inclusion in the target comparison set. In FIG. 11, the comparison set is illustrated as 1106 and the target comparison set is illustrated as 1108. For illustration purposes, the target comparison set has six members, with IDs ranging from A . . . G as illustrated by 1110. Thus A . . . G are those items with scores above average average of the reliance score of the set of user keys. If the primary metric is to increase the scores in the top 10 percentile, then 1110 would be those scores in the top 10 percentile, and so forth.

Operation 1112 represents selecting the exemplary model. The first operation in selecting the exemplary model is typically ordering the target comparison set by the second metric as indicated by operation 1114. Since a secondary improvement metric need not be selected in all instances, if there is no secondary metric, the system can apply a default secondary metric, a default ordering criteria, and/or a default selection criteria to select the exemplary model. In an example embodiment, when no secondary metric has been selected, increasing the overall average reliance score is used as a default secondary metric.

In FIG. 11, if the secondary improvement metric is to lower cost, then 1116 illustrates the target comparison set ordered by cost (high to low in this instance although low to high would work equally well). When this ordering takes place, multiple items may have the same value. Thus, G and C have the same cost and A and F are illustrated as having the same cost.

Operation 1118 then selects the appropriate item or items based on the secondary metric. Thus, if the secondary goal was to lower cost, and item D had the lowest cost of the target comparison set, then item D would be selected as the exemplary model as illustrated by 1124.

If only one item is to be used as an exemplary model and there are two or more items, then some tie-breaking criteria can be used to select between the choices. Although any tie breaking criteria can be used, in some embodiments another secondary or primary metric can be the tie breaker. By way of example, and not limitation, if the primary metric was to increase average score, and the secondary metric was to lower cost, and two items had the lowest cost, the one with the highest score could be the tie breaker. If the primary metric was to decrease the percentage of items in the lowest percentile and the secondary metric was to use the most common set of attributes, the highest score or lowest cost could be used as a tie-breaker.

In some embodiments it is allowable to select more than one exemplary model by taking the top/bottom n items. Say for example, the secondary metric was to increase some metric and items G, C, and E represented the top metrics, if the system was set up to take the top three items, then items G, C and E would all be chosen to make up the exemplary models.

Although operations 1114 and 1118 are indicated as first ordering the set 1110 and then selecting one or more items out of the set, those of skill in the art will understand that ordering first may not be required in all instances. For example, looping over all entries and selecting n entries with the highest or lowest metric without first ordering the metrics can be used in some embodiments.

FIG. 12 illustrates a flow diagram that gives three examples of primary improvement metrics: 1) move scores into a designated percentile, such as increasing the number of scores in the top 10 percentile; 2) moving scores out of a designated percentile, such as decreasing scores in the bottom 5 percentile; and 3) increasing/decreasing a metric, such as increasing the average score, decreasing some dispersion metric, etc. Thus, operation 1104 selects which primary improvement metric is indicated (operation 1202) and picks the correct methodology of selecting the target comparison set (operations 1204, 1206, 1208).

If the primary improvement metric is to increase the amount of the selected user key entities in a designated percentile (such as the top 10 percentile), the target comparison set consists of key entities of the comparison set with security reliance overall scores in the designated percentile (such as the top 10 percentile) as indicated in operation 1204.

If the primary improvement metric is to decrease the amount of the selected user key entities in a designated percentile (such as the bottom 5 percentile), the target comparison set consist of the key entities of the comparison set with security reliance overall scores in the complement of the designated percentile as indicated in operation 1206. Thus, if the improvement metric is to decrease the key entities in the bottom 5 percentile, the target comparison set is the top 95 percentile, the complement of the bottom 5 percentile.

If the primary improvement metric is to increase a metric such as the average security reliance overall score of the selected user key entities, the target comparison set consist of all key entities of the comparison set with a score that is above the average security reliance score as indicated by operation 1208.

In operation 1112, the target comparison set is subject to an optimizing calculation based on the selected secondary improvement metric of the improvement goal. The optimization process selects the exemplary model based on the secondary improvement metric. This is performed differently for different secondary improvement metrics. Operation 1210 selects the appropriate methodology for the given secondary improvement metric. FIG. 12 illustrates three different secondary improvement metrics: 1) lower cost; 2) increasing/decreasing a metric, such as increasing the average overall score; and 3) using the most common combination of attributes.

If the secondary improvement metric is to have lower costs, the key entities of the target comparison set are ordered by their respective costs as illustrated in operation 1212. The key entity with the lowest cost is then selected for the exemplary model with the highest overall reliance score acting as a tie-breaker between multiple key entities with the same minimal cost as indicated in operation 1214. Costs can either be a monetary cost or some other cost such as complexity/ease of implementation, time to implement, and so forth, or a combination of both.

If the secondary improvement metric is to increase/decrease a metric such as increasing the overall average security reliance score, the key entities of the target comparison set are ordered by the appropriate metric (i.e., their security reliance overall score) as illustrated in operation 1216. The key entity with the highest score (i.e., security reliance overall score) is then selected for the exemplary model with the lowest cost acting as a tie-breaker between multiple key entities with the same security reliance overall score as indicated in operation 1218.

If the secondary improvement metric is to match the most common attribute value combination, system counts the number of key entries with each attribute combination as indicated in operation 1220. In some embodiments, the key entities of the target comparison set are ordered by the frequency of attribute value combination. The key entity with most frequent attribute value combination is then selected for the exemplary model with the lowest cost acting as a tie-breaker between multiple key entities with the same occurrence count as indicated by operation 1222.

As one example, suppose the primary improvement metric is to increase the percentage of scores in the top 10 percentile range and the secondary improvement metric is to match the most common attribute value combination. Let's assume, the comparison set of cryptographic key material consists of TLS server certificates in the vertical industry “health care” and in the geographic region “north America.” Suppose the target comparison set, the top 10 percentile, consists of 100 entities with the following attribute value combination. KE_{1 . . . , 42}:{TLS Compression disabled, (Multiple) Certificate Status Request not supported, TLS Insecure Renegotiation disabled, TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA₂₅₆, SCT present, TLSA resource record not present, HSTS supported, HPKP not supported, certificate's cryptoperiod 2 years, sha256WithRSAEncryption, CDP II, OCSP KE_{43, . . . , 79}: {TLS Compression disabled, (Multiple) Certificate Status Request supported, TLS Insecure Renegotiation disabled, TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA₃₈₄, SCT present, TLSA resource record not present, HSTS supported, HPKP supported, certificate's cryptoperiod 1 years, ecdsa-with-SHA₃₈₄, CDP I, OCSP KE_{80, . . . , 100}:{TLS Compression disabled, (Multiple) Certificate Status Request supported, TLS Insecure Renegotiation disabled, TLS_RSA_WITH_AES_128_CBC_SHA₂₅₆, SCT not present, TLSA resource record not present, HSTS supported, HPKP supported, certificate's cryptoperiod 2 years, sha256WithRSAEncryption, CDP II, OCSP I}. Though the order of the average security reliance overall score for the given attribute value combination is σ(KE_{43, . . . , 79})>σ(KE_{1, . . . , 42})>σ(KE_{80, . . . , 100}), the secondary improvement metric, “match the most common attribute value combination”, mandates the exemplary model to pick the attribute value combination of KE_{1, . . . , 42}, which has 42 entries as compared to KE_{43, . . . , 79}, which has 37 entries and KE_{80, . . . , 100}, which has 21 entries. The most common attribute set can be identified by looping over all the various entities in the target comparison set and counting the number of entities with each of the different attribute combinations.

Returning to operation 1006 of FIG. 10, once the model has been created, the improvement potential is calculated in operation 1008. Improvement potential can be based on a variety of different strategies, all of which will result in improvement in some sense. As discussed above, the user may have a particular improvement goal, such as increasing the average security reliance score while minimizing the associated costs, increasing the percentage in the top percentile while matching the most common attribute value combination, decreasing the percentage in the bottom percentile while increasing the average security reliance overall score and so forth. To accomplish these improvement goals, a variety of strategies resulting in actions applicable to the selected user cryptographic key material may be employed. The strategies involve changing at least some cryptographic key material in the user set of cryptographic key material from their existing attribute configuration to the attribute configuration of the exemplary model. This may mean changing specific attributes of cryptographic key material from one value to another, reconfiguring systems, and so forth.

In one embodiment a recommended action that results in increasing the average security reliance overall score while minimizing the associated costs is achieved through the “replacement” of selected certificates (or other cryptographic material) with new instances that have the attributes of the exemplary model. For specific attributes, this would amount to recommending an adjustment from some existing configuration to an exemplary attribute value. For example, if several key entities of the sample subset have the same associated security reliance overall score, the key entity, after breaking a possible tie as described above, with the lowest associated cost value is picked for the exemplary model. Suppose, the attribute “cryptoperiod” in the model was “one year cryptoperiod,” then a corresponding improvement action can be defined by replacing those certificates with a cryptoperiod of more than one year with a cryptoperiod value of “one year cryptoperiod.” Thus, the recommendation would be to adjust the attribute “cryptoperiod” from the value “two year cryptoperiod” to an exemplary value “one year cryptoperiod”.

In another embodiment the recommended action to increase the average security reliance overall score while matching the most common attribute value combination may be achieved through a “reconfiguration” of servers that employ TLS server certificates selected by the user according to the corresponding attributes in the exemplary model. For specific attributes, this would amount to recommending an adjustment to an exemplary attribute value, e.g., the recommendation for the property “TLS configuration” could be the exemplary attribute “Disable TLS Insecure Renegotiation” and “Support HSTS”, if these match the most common attribute values in the exemplary model.

In yet another embodiment a recommended action for decreasing the proportion of cryptographic material in the lowest percentile while increasing the average security reliance overall score is by replacing keys in the lowest percentile with keys having attributes of the exemplary model. For example, when considering the SSH keys selected by the user in the lowest percentile range of a chosen sample subset is achieved through “rotation” of the selected SSH keys according to the corresponding attributes in the exemplary model. For specific attributes, this would amount to recommending an adjustment to an exemplary attribute value, e.g., the key entities of the sample subset's complement percentiles might encompass security strengths of {192, 256} bits for the attribute “key size” in which case the recommendation could be to increase the size of newly generated keys to meet a security strength of 256 bits.

Returning to operation 1008, the improvement potential for the selected user's cryptographic material can be calculated by looking at the impact that the adjustments above would have on the statistics/metrics presented to the user. As the system identifies actions (discussions below), the impact of the action on the primary or primary and secondary metrics can be calculated should the action be taken. For example, if the primary improvement metric aims at increasing the proportion of the selection of cryptographic keys in the top percentile range of the security reliance overall score while increasing the average security reliance overall score, i.e., the secondary improvement metric, are respectively computed for both the presence and the absence of the recommended improvement actions. The difference between these two metric values populates the corresponding “Primary improvement impact” and “Secondary improvement impact” placeholders (i.e., 926, 928, 930, 932). For example, let three of m selected TLS server certificates belong to the second best security reliance overall score percentile range. Let the recommended action be “replacement” of the selected certificates by new certificates adhering to the attributes of the exemplary model certificate. In this case, the proportion of cryptographic keys in the top percentile range will increase by 3/N where N is the number of cryptographic keys for which the user is responsible. This increase populates the “Primary improvement impact” placeholder. Suppose, the average security reliance overall score of the m TLS server certificates was x and the security reliance score for the exemplary model certificate is y, then the “Secondary improvement impact” placeholder is populated with (y−x)/m.

In addition, if the associated cost of applying a recommended action is known, e.g., by a user configuration, or can be derived by querying public resources, e.g., the different prices for TLS server certificates issued by a public CA, the estimated additional cost per cryptographic key and the total additional cost for all selected cryptographic key entries is calculated and displayed, see FIG. 9.

For example, let the recommended action to decrease the proportion of the selected certificate in the lowest percentile range of the security reliance overall score be the upgrade of domain-vetted (DV) certificates, priced by the previously issuing public CA, CA₁ at $c₁ per certificate, to extended-validation (EV) certificates, priced by the lowest charging public CA, CA₂ at $c₂, c₂>c₁. The estimated additional cost per certificate for applying this action would be $(c₂−c₁) per certificate and for n selected certificates the total additional cost would amount to $n·(c₂−c₁).

In another example, if one action is to replace/rotate key material having certain attribute values with model attribute values, then the statistics/metrics can be recalculated as if the user had chosen the replacement/rotation option. The difference between the existing statistics/metrics and the hypothetical statistics/metrics represents the improvement potential of that action. Similarly, if the action is a reconfiguration using model configuration attribute values, then the statistics/metrics can be recalculated as if the user had chosen the reconfiguration option. The difference between the existing statistics/metrics and the hypothetical statistics/metrics represents the improvement potential of that action.

In order to select which options to present to the user as possible recommendations, the system can calculate various combinations and present only those options that meet certain criteria. For example, if the user's improvement goal is to reduce the percentage of scores in the lowest percentile while increasing the average security reliance overall score, and based on the exemplary model the system determines that this can be accomplished by replacing certain certificates with certain model attributes, by reconfiguring the system, or both, the system may compare the various combinations and present only those choices that result in a designated improvement. Thus, if the user only wants to see choices that reduce the percentage of scores in the lowest percentile to 5% or less, the system can present only choices that meet the criteria.

In some instances, there may be many more choices than a user will want to consider even when filtering by criteria such as those above. In such an instance, the system may use further criteria to reduce the choices presented such as the choices that result in the fewest certificates replaced/rotated, the fewest attributes changed, the fewest reconfigurations, the fewest systems involved, and/or so forth. These examples are based on the assumption that the more changes that occur, the more costs that are incurred. Furthermore, if the system knows specific costs or relative costs (i.e., making a change to this system is twice as expensive as making a change to these other systems), the system can factor these in so as to minimize costs. In this context cost may be in dollars, time, complexity or any other such measure.

The foregoing may be performed by using various techniques such as calculating the improvement potential for various changes and then selecting those that meet specified goal(s)/criteria and then taking the top N choices for display. Other algorithms for “optimization” can be employed such as looking at which changes give the most improvement and then selecting those with the lowest cost, or within a pre-defined budget or any other such techniques.

Turning for a moment to FIG. 13, an example of how a set of actions can be identified is presented. The method, shown generally as 1300 takes as an input the item(s) identified as the exemplary model 1302. In FIG. 13, exemplary model 1302 is shown as having five attributes, along with their corresponding values 1, 2, 3 4, and 5. In case multiple exemplary models have been identified, each of these models gives rise to a distinct set of recommended actions. The other input is the set of user keys to be improved 1304. In FIG. 13, this is represented by U₁, . . . U_n, along with the corresponding attributes and values.

The method then compares the attribute values of the exemplary model(s) 1302 with the attribute values of the set 1304 and identifies transformations that can be taken to convert the attribute values of set 1304 into the attribute values of the exemplary model(s) 1302. In FIG. 13 the identified transformations are represented by 1306. The transformations are specified by T₁, T₂, etc. Where attribute values of 1304 already match the attribute values of the exemplary model(s) 1302, then no transformation need be taken (represented in FIG. 13 by a simple “X”). Once the necessary transformations are identified for a user key they are assembled to transformations sets, specified by TS₁, TS₂, etc., as illustrated in 1310.

Transformations, illustrated as 1316, are deterministically mapped to operations, specified by O₁, O₂, etc. illustrated as 1318, which are actionable and usually proprietarily defined by a key management system processing the user's keys. This mapping can be viewed as a many-to-many relationship, i.e., several transformations may be mapped to a single operation (e.g., T1 and T3 are mapped to O2) or a single transformation may be mapped to several operations (e.g., T2 is mapped to both O1 and O3). This mapping is based on what operation(s) are performed to accomplish the identified transformation and include such operations as key rotation, certificate re-issue, system (re)configuration, and so forth.

The many-to-many mapping can result in a transformation set being mapped to alternative actions. For example, in FIG. 13, to transform user key U₂ into the exemplary model the second attribute has to be transformed from value 8 to value 2 and the forth attribute has to be transformed from attribute value 9 to attribute value 4. These transformations are illustrated by T₂ and T₄ respectively, so transformation set TS₂ is the set {T₂, T₄}. The mapping of 1316 to 1318 shows that T₂ can be accomplished either by operation O₁ or by operation O₃ and that T₄ can be accomplished by operation O₁. Thus, to accomplish the transformation, there are two alternative actions, A₂, consisting of operations O₁ and O₃ and A₃ consisting of operation O₁. Either of these actions will accomplish the desired transformation.

Based on this mapping, actions, specified by A₁, A₂, etc., are created as sets of those operations, whose transformations constitute the respective transformation set. Actions are then applicable to a subset of the user's key selection and may be shown to the user as illustrated in FIG. 9, i.e., 926, 928, 930, and 932, or in a non-interactive mode been automatically executed as described in more detail below.

For example, assuming an exemplary model EM for SSH key material and a user key U₁ consist of attribute values

			Best supported
	Protocol version	Key exchange algorithms	cipher alg.	Auth. method
EM	2.0	curve25519-sha256@libssh.org,	aes-256-cbc	publickey
		ecdh-sha2-nistp256,
		ecdh-sha2-nistp384,
		ecdh-sha2-nistp521,
		diffie-hellman-group-exchange-sha256,
		diffie-hellman-group14-sha1
U1	2.0	diffie-hellman-group-exchange-sha256,	aes-256-cbc	publickey
		diffie-hellman-group14-sha1,
		diffie-hellman-group-exchange-sha1,
		diffie-hellman-group1-sha1

Then the necessary transformation can be defined as T₁:=“Include support for curve25519-sha256@libssh.org, ecdh-sha2-nistp256,ecdh-sha2-nistp384,ecdh-sha2-nistp521; Exclude support for diffie-hellman-group-exchange-sha1, diffie-hellman-group1-sha1”. The transformation set TS₁ consists of this transformation only, i.e., TS₁: ={T₁}. The transformation for T₁ may be mapped to the operation O₁₇:=“sshd re-configuration” which in this case is parametrized by “set KexAlgorithms to {curve25519-sha256@libssh.org, ecdh-sha2-nistp256,ecdh-sha2-nistp384,ecdh-sha2-nistp521, diffie-hellman-group-exchange-sha256, diffie-hellman-group14-sha1}”. This leads to the action A₁: ={O₁₇} which may be shown to the user as “Reconfigure SSH server” in, say, panel 926.

In another, more complex, example, assuming an exemplary model EM for X.509 certificates and user keys U₁, . . . , U₄ consists of

					Certificate's
	Strongest supported cipher suite	SCT	PFS	Crypto-period	signature algorithm
EM	TLS_ECDHE_ECDSA_WITH_	Yes	Yes	1 year	sha256WithRSAEncryption
	AES_256_GCM_SHA384
U1	TLS_RSA_WITH_RC4_128_MD5	Yes	No	1 year	sha256WithRSAEncryption
U2	TLS_ECDHE_ECDSA_WITH_	No	Yes	2 years	sha256WithRSAEncryption
	AES_256_GCM_SHA384
U3	TLS_ECDHE_ECDSA_WITH_	Yes	Yes	1 year	shaWithRSAEncryption
	AES_256_GCM_SHA384
U4	TLS_ECDHE_ECDSA_WITH_	Yes	Yes	2 years	shaWithRSAEncryption
	AES_256_GCM_SHA384

Then the necessary transformations can be defined as
T₁: =“Include cipher suite TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA₃₈₄”,

T₂: =“Provide an SCT”,

T₃: =“Support PFS”,

T₄: =“Set certificate's validity period to 1-year”, and
T₅: =“Set certificate's signature algorithm to sha256WithRSAEncryption”.
The resulting transformation sets are,

• TS₁: ={T₁},
• TS₂: ={T₂, T₄},
• TS₃: ={T₅}, and
• TS₄: ={T₄, T₅}.
• The transformation mapping may be defined as
• T₁ O₂:=“Modify httpd configuration” parameterized by “include cipher suite TLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA₃₈₄”

$T_2\mapsto \{\begin{array}{cc}{O}_1\text{:=}& \begin{array}{c}“\mathrm{Enroll}\mathrm{for}\mathrm{certificate}”\mathrm{parametrized}\mathrm{by}\\ “\mathrm{type}\text{:}\mathrm{extended}\text{-}\mathrm{validation}\left(\mathrm{EV}\right)”\left[\mathrm{contains}\mathrm{embedded}\mathrm{SCTs}\right]\end{array}\\ O_3\text{:=}& \begin{array}{c}“\mathrm{Modify}\mathrm{TLS}\mathrm{extension}”\mathrm{parametrized}\mathrm{by}\\ “\mathrm{set}{\mathrm{signed}}_{—}{\mathrm{certificate}}_{—}\mathrm{timestamp}”\left[\mathrm{support}\mathrm{SCT}\mathrm{forwarding}\right]\end{array}\end{array}$

• T₃O₂:=“Modify httpd configuration” parameterized by “include at least one cipher suite with support of ephemeral Diffie-Hellman (DHE/ECDHE) key exchange”
• T₄O₁:=“Enroll for certificate” parametrized by “validity period: 1-year”
• T₅O₁:=“Enroll for certificate” parametrized by “signature algorithm: sha256WithRSAEncryption”
  The resulting actions, applicable to the corresponding subset of user keys, are
• A₁: ={O₂} which may be shown to the user as “Modify TLS server configuration”,
• A₂: ={O₁, O₃}, which may be shown to the user as “Modify TLS server configuration and replace certificate”, and
• A₃: ={O₁}, which may be shown to the user as “Replace certificate”.

If there are more actions than a user might want to see or more actions than a system presents, then the number of actions to be presented/used can be filtered in some fashion as described above and as illustrated by 1314.

Once the system identifies which actions to use (as, for example, set 1314), the system presents the choices as indicated in operation 1010. As indicated in conjunction with FIG. 9, the system may also have a user interface element that allows the user to kick off one or more processes to implement (or begin implementing) selected actions (i.e., 926, 928, 930, and 932). If the user selects such action(s), the system can respond by initiating the selection action(s) as illustrated in operation 1012.

For situations where the system is not being used interactively, as discussed below, the system may not display information as discussed above. Rather the system may use the calculated improvement potential (operation 1008) and the improvement potential and/or other criteria may be used to select an action in operation 1011. For example, the action(s) with the highest improvement potential may be selected or action(s) may be selected based on some other criteria. After an action is selected, the selected action may be initiated as indicated in operation 1012.

As mentioned above, the process of FIGS. 10-13 may be run in a non-interactive manner and thus may not present the user interface of FIG. 9 to a user and receive input thereby or output information thereto. Automated operation of the processes of FIGS. 10-13 may occur in a variety of contexts/embodiments. These can be based, for example, on particular events that kick off operation of the processes in FIGS. 10-13. The following represent examples of situations where the processes of FIGS. 10-13 can be used in an automated fashion. While they are representative in nature, they do not represent an exhaustive list.

In one situation, the system can have preselected sets of user cryptographic material that are monitored for particular events. As noted above, the security reliance score can change over time, such as through operation of score adjustment and the learning model(s) described above. The system can monitor various metrics about sets/subsets of user cryptographic material and when certain events occur, trigger the processes in FIGS. 10-13 to automatically adjust the attributes of cryptographic key material. For example, a particular set/subset may be monitored and when the overall score drops into a particular target percentile, relative to some comparison set of cryptographic material, corrective action can be taken. In another example, the average security reliance overall score for a particular set/subset may be monitored and compared against a threshold and when the average score transgresses the threshold, corrective action can be taken. In yet another example, some sort of debasing criteria is met. As an example of this last type of event, if, for example, a debasing reliance score re-evaluation or hitherto unknown vulnerability is discovered affecting a particular attribute/configuration, a system administrator may want to automatically take corrective action, say by replacing compromised or potentially compromised keys with a hitherto sufficient but now as weak considered security strength or reconfigure systems that use a particular, now vulnerable configuration. When any of these events occur, corrective action can be taken through the processes of FIGS. 10-13. In a further example, the processes of FIGS. 10-13 can be run according to a schedule (i.e., periodically or aperiodically) and the actions taken automatically as described above. Combinations thereof are also within the scope of the invention. Thus, the occurrence of an event can trigger operation of the processes of FIGS. 10-13 on a schedule or the occurrence of an event can end operation of the processes of FIGS. 10-13 on a schedule or any other combination of one or more schedules and and one or more event based operation. Multiple schedules can also be used in some embodiments.

An example can help illustrate how this can all occur. In this example, the system monitors a particular set of user cryptographic key material for the event that the percentage of cryptographic key material in the bottom 5 percentile exceeds 10 percent. The improvement goal in this example is set by an administrator to be to reduce the number of cryptographic key material in the bottom 5 percentile while using the most common set of attributes. Thus, in this example, the primary improvement metric (which is the same as the monitored event) is to reduce the number of cryptographic key material in the bottom 5 percentile and the secondary improvement metric is to use the most common attribute set.

When the triggering event occurs, the process of FIG. 10 is started and operation 1002 retrieves the set of user cryptographic key material. To the extent that statistics/metrics are used (i.e., to calculate improvement potential) they can be calculated in operation 1004. The exemplary model is then created in operation 1006 as illustrated by the process in FIGS. 11-12. In this example, operation 1104 will select the target comparison set as the remaining 95 percentile of the comparison set. Since the secondary improvement metric is using the most common attributes, operations 1220 and 1222 are executed and the the key material from the target comparison set with the most common combination of attributes is selected as the exemplary model.

The improvement potential is calculated in operation 1008 and operation 1011 selects an action, based in improvement potential, and any policies or metrics, as discussed above. Finally, the selected actions are initiated in operation 1012.

Logical System Architecture

FIG. 14 illustrates an example logical system architecture 1400. Such a logical architecture comprises various modules, such as analytics module 1402, scoring module 1404 and scoring aggregation module 1406 implemented as part of a compute cluster 1408 or other machine (not shown).

Analytics module 1402, for example, performs various operations such as the learning process, statistical sampling and other analytic aspects described above. Scoring module 1404, for example, calculates sub-scores as described above and scoring aggregation module 1406 aggregates individual scores into those described elsewhere. Other modules may include reporting modules, modules to calculate new factors, and so forth.

Computer cluster 1408 represents a location to implement the modules and logic described above. It can be, for example, the systems illustrated in FIG. 3 (e.g., 308) and/or FIG. 4 (e.g., 402).

Also illustrated are persistence services module 1410 which can store data in various databases such as data store 1412 and data store 1414. Two data stores are illustrated in order to represent that multiple levels of storage may be maintained, such as more immediate storage and more archival storage. ETL (Export Transform Load) Services module, in conjunction with specified data sources (such as the illustrated scanners, data feeds, export services 1418) provide the ability to get data into or out of the system in various ways. The ETL may be used, for example, for bulk export/import of information. Smaller amounts of information can use the client/API Reports interface 1420. The system may also provide an API or other mechanism for a client or other system to access the functionality provided by the system (1420). Such would be used, for example, by the described interactive tool or by another system to produce reports and so forth. The scheduling module provides scheduling services so that surveys, data gathering and so forth can be performed on a periodic basis according to a designated schedule. Other modules may also be implemented, although they are not specifically illustrated in FIG. 14.

Modules, Components and Logic

Certain embodiments are described herein as including logic or a number of components, modules, or mechanisms. Modules may constitute either software modules (i.e., code embodied on a machine-readable medium) or hardware-implemented modules. A hardware-implemented module is tangible unit capable of performing certain operations and may be configured or arranged in a certain manner. In example embodiments, one or more computer systems (e.g., a standalone, client or server computer system) or one or more processors may be configured by software (e.g., an application or application portion) as a hardware-implemented module that operates to perform certain operations as described herein.

In various embodiments, a hardware-implemented module may be implemented mechanically or electronically. For example, a hardware-implemented module may comprise dedicated circuitry or logic that is permanently configured (e.g., as a special-purpose processor, such as a field programmable gate array (FPGA) or an application-specific integrated circuit (ASIC)) to perform certain operations. A hardware-implemented module may also comprise programmable logic or circuitry (e.g., as encompassed within a general-purpose processor or other programmable processor) that is temporarily configured by software to perform certain operations. It will be appreciated that the decision to implement a hardware-implemented module mechanically, in dedicated and permanently configured circuitry, or in temporarily configured circuitry (e.g., configured by software) may be driven by cost and time considerations.

Accordingly, the term “hardware-implemented module” should be understood to encompass a tangible entity, be that an entity that is physically constructed, permanently configured (e.g., hardwired) or temporarily or transitorily configured (e.g., programmed) to operate in a certain manner and/or to perform certain operations described herein. Considering embodiments in which hardware-implemented modules are temporarily configured (e.g., programmed), each of the hardware-implemented modules need not be configured or instantiated at any one instance in time. For example, where the hardware-implemented modules comprise a general-purpose processor configured using software, the general-purpose processor may be configured as respective different hardware-implemented modules at different times. Software may accordingly configure a processor, for example, to constitute a particular hardware-implemented module at one instance of time and to constitute a different hardware-implemented module at a different instance of time.

Hardware-implemented modules can provide information to, and receive information from, other hardware-implemented modules. Accordingly, the described hardware-implemented modules may be regarded as being communicatively coupled. Where multiple of such hardware-implemented modules exist contemporaneously, communications may be achieved through signal transmission (e.g., over appropriate circuits and buses) that connect the hardware-implemented modules. In embodiments in which multiple hardware-implemented modules are configured or instantiated at different times, communications between such hardware-implemented modules may be achieved, for example, through the storage and retrieval of information in memory structures to which the multiple hardware-implemented modules have access. For example, one hardware-implemented module may perform an operation, and store the output of that operation in a memory device to which it is communicatively coupled. A further hardware-implemented module may then, at a later time, access the memory device to retrieve and process the stored output. Hardware-implemented modules may also initiate communications with input or output devices, and can operate on a resource (e.g., a collection of information).

The various operations of example methods described herein may be performed, at least partially, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors may constitute processor-implemented modules that operate to perform one or more operations or functions. The modules referred to herein may, in some example embodiments, comprise processor-implemented modules.

Similarly, the methods described herein are at least partially processor-implemented. For example, at least some of the operations of a method may be performed by one or more processors or processor-implemented modules. The performance of certain of the operations may be distributed among the one or more processors, not only residing within a single machine, but deployed across a number of machines. In some example embodiments, the processor or processors may be located in a single location (e.g., within a home environment, an office environment or as a server farm), while in other embodiments the processors may be distributed across a number of locations.

The one or more processors may also operate to support performance of the relevant operations in a “cloud computing” environment or as a “software as a service” (SaaS). For example, at least some of the operations may be performed by a group of computers (as examples of machines including processors), these operations being accessible via a network (e.g., the Internet) and via one or more appropriate interfaces (e.g., application program interfaces (APIs).)

Electronic Apparatus and System

Example embodiments may be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or in combinations of them. Example embodiments may be implemented using a computer program product, e.g., a computer program tangibly embodied in an information carrier, e.g., in a machine-readable medium for execution by, or to control the operation of, data processing apparatus, e.g., a programmable processor, a computer, or multiple computers.

A computer program can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, subroutine, or other unit suitable for use in a computing environment. A computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a communication network.

In example embodiments, operations may be performed by one or more programmable processors executing a computer program to perform functions by operating on input data and generating output. Method operations can also be performed by, and apparatus of example embodiments may be implemented as, special purpose logic circuitry, e.g., a field programmable gate array (FPGA) or an application-specific integrated circuit (ASIC).

The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. In embodiments deploying a programmable computing system, it will be appreciated that that both hardware and software architectures may be employed. Specifically, it will be appreciated that the choice of whether to implement certain functionality in permanently configured hardware (e.g., an ASIC), in temporarily configured hardware (e.g., a combination of software and a programmable processor), or a combination of permanently and temporarily configured hardware may be a design choice. Below are set out hardware (e.g., machine) and software architectures that may be deployed, in various example embodiments.

Example Machine Architecture and Machine-Readable Medium

FIG. 15 is a block diagram of a machine in the example form of a processing system within which may be executed a set of instructions for causing the machine to perform any one or more of the methodologies discussed herein including the functions, systems and flow diagrams thereof.

In alternative embodiments, the machine operates as a standalone device or may be connected (e.g., networked) to other machines. In a networked deployment, the machine may operate in the capacity of a server or a client machine in server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine may be a personal computer (PC), a tablet PC, a set-top box (STB), a personal digital assistant (PDA), a cellular telephone, a smart phone, a tablet, a wearable device (e.g., a smart watch or smart glasses), a web appliance, a network router, switch or bridge, or any machine capable of executing instructions (sequential or otherwise) that specify actions to be taken by that machine. Further, while only a single machine is illustrated, the term “machine” shall also be taken to include any collection of machines that individually or jointly execute a set (or multiple sets) of instructions to perform any one or more of the methodologies discussed herein.

The example of the machine 1500 includes at least one processor 1502 (e.g., a central processing unit (CPU), a graphics processing unit (GPU), advanced processing unit (APU), or combinations thereof), a main memory 1504 and static memory 1506, which communicate with each other via bus 1508. The machine 1500 may further include graphics display unit 1510 (e.g., a plasma display, a liquid crystal display (LCD), a cathode ray tube (CRT), and so forth). The machine 1500 also includes an alphanumeric input device 1512 (e.g., a keyboard, touch screen, and so forth), a user interface (UI) navigation device 1514 (e.g., a mouse, trackball, touch device, and so forth), a storage unit 1516, a signal generation device 1528 (e.g., a speaker), sensor(s) 1521 (e.g., global positioning sensor, accelerometer(s), microphone(s), camera(s), and so forth) and a network interface device 1520.

Machine-Readable Medium

The storage unit 1516 includes a machine-readable medium 1522 on which is stored one or more sets of instructions and data structures (e.g., software) 1524 embodying or utilized by any one or more of the methodologies or functions described herein. The instructions 1524 may also reside, completely or at least partially, within the main memory 1504, the static memory 1509, and/or within the processor 1502 during execution thereof by the machine 1500. The main memory 1504, the static memory 1509 and the processor 1502 also constituting machine-readable media.

While the machine-readable medium 1522 is shown in an example embodiment to be a single medium, the term “machine-readable medium” may include a single medium or multiple media (e.g., a centralized or distributed database, and/or associated caches and servers) that store the one or more instructions or data structures. The term “machine-readable medium” shall also be taken to include any tangible medium that is capable of storing, encoding or carrying instructions for execution by the machine and that cause the machine to perform any one or more of the methodologies of the present invention, or that is capable of storing, encoding or carrying data structures utilized by or associated with such instructions. The term “machine-readable medium” shall accordingly be taken to include, but not be limited to, solid-state memories, and optical and magnetic media. Specific examples of machine-readable media include non-volatile memory, including by way of example semiconductor memory devices, e.g., erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), and flash memory devices; magnetic disks such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks. The term machine-readable medium specifically excludes non-statutory signals per se.

Transmission Medium

The instructions 1524 may further be transmitted or received over a communications network 1526 using a transmission medium. The instructions 1524 may be transmitted using the network interface device 1520 and any one of a number of well-known transfer protocols (e.g., HTTP). Transmission medium encompasses mechanisms by which the instructions 1524 are transmitted, such as communication networks. Examples of communication networks include a local area network (“LAN”), a wide area network (“WAN”), the Internet, mobile telephone networks, plain old telephone (POTS) networks, and wireless data networks (e.g., WiFi and WiMax networks). The term “transmission medium” shall be taken to include any intangible medium that is capable of storing, encoding or carrying instructions for execution by the machine, and includes digital or analog communications signals or other intangible media to facilitate communication of such software.

Although an embodiment has been described with reference to specific example embodiments, it will be evident that various modifications and changes may be made to these embodiments without departing from the broader spirit and scope of the invention. Accordingly, the specification and drawings are to be regarded in an illustrative rather than a restrictive sense. The accompanying drawings that form a part hereof, show by way of illustration, and not of limitation, specific embodiments in which the subject matter may be practiced. The embodiments illustrated are described in sufficient detail to enable those skilled in the art to practice the teachings disclosed herein. Other embodiments may be utilized and derived therefrom, such that structural and logical substitutions and changes may be made without departing from the scope of this disclosure. This Detailed Description, therefore, is not to be taken in a limiting sense, and the scope of various embodiments is defined only by the appended claims, along with the full range of equivalents to which such claims are entitled.

Such embodiments of the inventive subject matter may be referred to herein, individually and/or collectively, by the term “invention” merely for convenience and without intending to voluntarily limit the scope of this application to any single invention or inventive concept if more than one is in fact disclosed. Thus, although specific embodiments have been illustrated and described herein, it should be appreciated that any arrangement calculated to achieve the same purpose may be substituted for the specific embodiments shown. This disclosure is intended to cover any and all adaptations or variations of various embodiments. Combinations of the above embodiments, and other embodiments not specifically described herein, will be apparent to those of skill in the art upon reviewing the above description.

Claims

1. A method for improving security reliance scores of cryptographic key material comprising:

obtaining a set of user cryptographic key material and a set of comparison cryptographic key material, each cryptographic key material in the respective sets having an associated security reliance score based on attributes of the cryptographic key material;

identifying an improvement goal comprising a primary improvement metric and a secondary improvement metric;

creating an exemplary model cryptographic key material by performing operations comprising: based on the primary improvement metric, selecting a target comparison set of cryptographic key material to use as the basis for a model cryptographic key material; based on the secondary improvement metric, selecting at least one cryptographic key material in the target comparison set of cryptographic key material as the exemplary model cryptographic key material such that the attributes of the exemplary model cryptographic key material comprise the attributes of the selected at least one cryptographic key material;

selecting a subset of user cryptographic key material for improvement;

calculating improvement potential by performing operations comprising: calculating the primary improvement metric for the set of user cryptographic key material; creating a hypothetical set of user cryptographic key material by replacing the selected subset of user cryptographic key material with cryptographic key material having attributes of the exemplary model cryptographic key material; calculating the primary improvement metric for the hypothetical set of user cryptographic key material; using as the improvement potential the difference between the primary improvement metric for the hypothetical set of user cryptographic key material and the corresponding primary improvement metric for the set of user cryptographic key material; and

initiating an improvement action to realize at least a portion of the improvement potential.

2. The method of claim 1, wherein the primary improvement metric comprises at least one of:

moving cryptographic key material into a percentile;

moving cryptographic key material out of a percentile; and

increasing or decreasing an identified metric.

3. The method of claim 2 wherein the identified metric comprises at least one of:

an average security reliance score for the set of user cryptographic key material;

a median security reliance score for the set of user cryptographic key material;

a maximum security reliance score for the set of user cryptographic key material;

a minimum security reliance score for the set of user cryptographic key material; and

a dispersion metric.

4. The method of claim 1 wherein the secondary improvement metric comprises at least one of:

lower cost;

increasing or decreasing a secondary metric; and

most common attributes.

5. The method of claim 1 further comprising:

obtaining a set of comparison cryptographic key material;

calculating a comparison metric for the set of comparison cryptographic key material; and

presenting, as part of a user interface, the comparison metric along with a metric calculated for the set of user cryptographic key material.

5. The method of claim 1 wherein a second improvement metric is used to select the one of cryptographic key material in the target comparison set of cryptographic key material as the exemplary model cryptographic key material when the secondary improvement metric selects multiple cryptographic key material.

6. The method of claim 1, wherein the improvement goal comprises at least one of:

increasing a number of cryptographic key material in a designated percentile with lower cost;

increasing the number of cryptographic key material in the designated percentile with most common attributes;

increasing the number of cryptographic key material in the designated percentile while increasing a metric of the user set of cryptographic key material;

increasing the number of cryptographic key material in the designated percentile while decreasing a metric of the user set of cryptographic key material;

decreasing the number of cryptographic key material in a second designated percentile with lower cost;

decreasing the number of cryptographic key material in the second designated percentile with most common attributes;

decreasing the number of cryptographic key material in a designated percentile while increasing the metric of the user set of cryptographic key material;

decreasing the number of cryptographic key material in a designated percentile while decreasing the metric of the user set of cryptographic key material;

increasing the metric of the user set of cryptographic key material with lower cost;

decreasing the metric of the user set of cryptographic key material with lower cost;

increasing the metric of the user set of cryptographic key material with most common attributes; and

decreasing the metric of the user set of cryptographic key material with most common attributes.

7. The method of claim 1, further comprising:

identifying the occurrence of an event;

responsive to the occurrence of the event, performing the operations of claim 1.

8. The method of claim 7 wherein the occurrence of the event comprises a time of a schedule.

9. The method of claim 4, wherein the cost comprises at least one of: a monetary cost, a metric indicating complexity to implement, and a metric indicating time to implement.

10. The method of claim 1 further comprising:

presenting a user interface to a user, the user interface comprising at least one user interface control allowing a user to select the set of user cryptographic key material;

presenting to the user via the user interface, the calculated improvement potential along with at least one action, the at least one action including the improvement action; and

receiving via the user interface, user selection of the improvement action.

11. A machine-readable medium having executable instructions encoded thereon, which, when executed by at least one processor of a machine, cause the machine to perform operations comprising:

obtain a set of user cryptographic key material and a set of comparison cryptographic key material, each cryptographic key material in the respective sets having an associated security reliance score based on attributes of the cryptographic key material;

identify an improvement goal comprising a primary improvement metric and an optional secondary improvement metric;

create an exemplary model cryptographic key material by performing operations comprising: based on the primary improvement metric, select a target comparison set of cryptographic key material to use as the basis for a model cryptographic key material; when the improvement goal includes the optional secondary improvement metric, based on the optional secondary improvement metric select at least one cryptographic key material in the target comparison set of cryptographic key material as the exemplary model cryptographic key material; when the improvement goal does not include the optional secondary improvement metric, selecting a default secondary improvement metric and based on the default secondary improvement metric select at least one cryptographic key material in the target comparison set of cryptographic key material as the exemplary model cryptographic key material;

selecting a subset of user cryptographic key material for improvement;

calculating improvement potential by performing operations comprising: calculating the primary improvement metric for the set of user cryptographic key material; creating a hypothetical set of user cryptographic key material by replacing the selected subset of user cryptographic key material with cryptographic key material having attributes of the exemplary model cryptographic key material; calculating the primary improvement metric for the hypothetical set of user cryptographic key material; using as the improvement potential the difference between the primary improvement metric for the hypothetical set of user cryptographic key material and the corresponding primary improvement metric for the set of user cryptographic key material; and

initiate an improvement action to realize at least a portion of the improvement potential.

12. The machine-readable medium of claim 11, wherein the operations further comprise:

identify a set of improvement actions by: compare an attribute of the at least one exemplary model to a corresponding attribute in an entry in the subset of user cryptographic key material selected for improvement; add to the set of improvement actions an action that changes the attribute of the entry to have a value equivalent to a value of the attribute of the at least one exemplary model.

13. The machine-readable medium of claim 11 wherein the primary improvement metric comprises at least one of:

moving cryptographic key material into a percentile;

moving cryptographic key material out of a percentile; and

increasing or decreasing an identified metric.

14. The machine-readable medium of claim 13 wherein the identified metric comprises at least one of:

an average security reliance score for the set of user cryptographic key material;

a median security reliance score for the set of user cryptographic key material;

a maximum security reliance score for the set of user cryptographic key material;

a minimum security reliance score for the set of user cryptographic key material; and

a dispersion metric.

15. The machine-readable medium of claim 11 wherein the optional secondary improvement metric comprises at least one of:

lower cost;

increasing or decreasing a secondary metric; and

most common attributes.

16. The machine-readable medium of claim 11 wherein the default secondary improvement metric comprises increasing an average security reliance score.

17. A system comprising:

a processor and executable instructions accessible on a machine-readable medium that, when executed, cause the system to perform operations comprising: obtain a set of user cryptographic key material and a set of comparison cryptographic key material, each cryptographic key material in the respective sets having an associated security reliance score based on attributes of the cryptographic key material; identify an improvement goal comprising a primary improvement metric and an optional secondary improvement metric; create an exemplary model cryptographic key material by performing operations comprising: based on the primary improvement metric, select a target comparison set of cryptographic key material to use as the basis for a model cryptographic key material; when the improvement goal includes the optional secondary improvement metric, based on the optional secondary improvement metric select at least one cryptographic key material in the target comparison set of cryptographic key material as the exemplary model cryptographic key material; when the improvement goal does not include the optional secondary improvement metric, selecting a default secondary improvement metric and based on the default secondary improvement metric select at least one cryptographic key material in the target comparison set of cryptographic key material as the exemplary model cryptographic key material; selecting a subset of user cryptographic key material for improvement; calculating improvement potential by performing operations comprising: calculating the primary improvement metric for the set of user cryptographic key material; creating a hypothetical set of user cryptographic key material by replacing the selected subset of user cryptographic key material with cryptographic key material having attributes of the exemplary model cryptographic key material; calculating the primary improvement metric for the hypothetical set of user cryptographic key material; using as the improvement potential the difference between the primary improvement metric for the hypothetical set of user cryptographic key material and the corresponding primary improvement metric for the set of user cryptographic key material; and initiate an improvement action to realize at least a portion of the improvement potential.

18. The system of claim 17, wherein the operations further comprise:

present a user interface to a user, the user interface comprising at least one user interface control allowing a user to select the set of user cryptographic key material;

present to the user via the user interface, the calculated improvement potential along with at least one action, the at least one action including the improvement action; and

receive via the user interface, user selection of the improvement action.

19. The system of claim 18 wherein the operations further comprise:

identifying the occurrence of an event;

responsive to the occurrence of the event, performing the operations of claim 17.

20. The system of claim 19, wherein the occurrence of the event comprises a time of a schedule.